Twins by content: at most 500 twins read per pair

With the hash typed as the column (f537add) the planner walks every twin of a
pair by index; 55 pairs have over 500 twins, up to 131k, and a flush with such
a pair ran the query for 37 s inside its transaction, the other flushes and
node-api's class counters waiting on it. A LATERAL per pair now reads its
lowest 500 release ids and takes the first named one: 49 ms for 305 pairs with
the five hottest among them. LOG.md (138).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

.gitea/workflows/deploy.yaml

@@ -59,35 +59,28 @@ jobs:
           node-version-file: .nvmrc
           cache: npm
       - run: npm ci
+      - name: Rust toolchain
+        run: |
+          command -v cargo || curl -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal
+          echo "$HOME/.cargo/bin" >> "$GITHUB_PATH"
+      # The edge worker (edge/, Rust): built first, since the e2e suite runs edge/target/release/edge-node against
+      # the node API; then its own tests, and the Rust and TypeScript RangeIndexers compared on the committed
+      # captures (edge/fixtures).
+      - name: Build the edge worker
+        run: cargo build --release --manifest-path edge/Cargo.toml --bin edge-node
       - name: Type check
         run: npx tsc --noEmit
       - name: Migrate
         run: node ace migration:run --force
       - name: Test
         run: node ace test
-
-  # The edge worker (edge/, Rust): its tests, and the Rust and TypeScript
-  # RangeIndexers compared on the committed captures (edge/fixtures).
-  edge:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-node@v4
-        with:
-          node-version-file: .nvmrc
-          cache: npm
-      - run: npm ci
-      - name: Rust toolchain
-        run: |
-          command -v cargo || curl -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal
-          echo "$HOME/.cargo/bin" >> "$GITHUB_PATH"
       - name: cargo test
         run: cargo test --release --manifest-path edge/Cargo.toml
-      - name: Rust and TypeScript identical on the fixtures
-        run: edge/check-fixtures.sh
+      - name: The Rust indexer over the fixtures
+        run: edge/bench-fixtures.sh
 
   deploy:
-    needs: [test, edge]
+    needs: [test]
     if: github.ref == 'refs/heads/rebuild'
     runs-on: ubuntu-latest
     steps:

Dockerfile

@@ -8,6 +8,15 @@
 # validate commands) throws "Invalid URL" on $ref, so no ace command starts.
 ARG NODE_VERSION=24.16
 
+# The Rust worker (before the app: its image carries edge-bench). Registry and target dir are build caches, so a rebuild only compiles what changed.
+FROM rust:1-bookworm AS edge-build
+WORKDIR /src
+COPY edge ./
+RUN --mount=type=cache,target=/usr/local/cargo/registry \
+    --mount=type=cache,target=/src/target \
+    cargo build --release --locked --bin edge-node --bin edge-bench \
+ && cp target/release/edge-node target/release/edge-bench /usr/local/bin/
+
 # The app, compiled, with production dependencies only.
 FROM node:${NODE_VERSION}-bookworm AS app-build
 WORKDIR /src
@@ -22,19 +31,12 @@ FROM node:${NODE_VERSION}-bookworm-slim AS app
 ENV NODE_ENV=production
 WORKDIR /app
 COPY --from=app-build /src/build ./
+# The Rust indexer's bench: `node ace replay:corpus` runs a group's corpus through it (the replay harness).
+COPY --from=edge-build /usr/local/bin/edge-bench /usr/local/bin/edge-bench
 USER node
 # Every service picks its own command, e.g. `node ace coordinator:run`.
 CMD ["node", "ace", "list"]
 
-# The Rust worker. Registry and target dir are build caches, so a rebuild only compiles what changed.
-FROM rust:1-bookworm AS edge-build
-WORKDIR /src
-COPY edge ./
-RUN --mount=type=cache,target=/usr/local/cargo/registry \
-    --mount=type=cache,target=/src/target \
-    cargo build --release --locked --bin edge-node --bin edge-bench \
- && cp target/release/edge-node target/release/edge-bench /usr/local/bin/
-
 FROM debian:bookworm-slim AS edge
 COPY --from=edge-build /usr/local/bin/edge-node /usr/local/bin/edge-node
 # edge-bench replays the corpus for the grouping sweeps (deploy/corpus_sweep.py).

REBUILD.md

@@ -60,9 +60,21 @@ most workers on batcave, a few on poweredge-kantoor with
 `edge:node --max-mbit` (e.g. 50 of its 160 Mbit/s ≈ 100–125 k articles/s;
 the cap is shared by the node's processes).
 
+Every group is tiled from article 1 (2026-10-28): one "ancient" range,
+status `ancient`, covers 1 … low watermark − 1, what had expired on the
+server before the group was planned. Never leased, fetched or filled —
+nothing selects it — and the status panel reports it apart ("Ancient: …
+articles below the low watermark"); migration 6700 added it for the
+groups planned before, `Planner.historyRows` for groups planned since.
+
 Also built (2026-10-09): the live tail (new articles planned every 2
 minutes from one LIST ACTIVE per name prefix, at priority −1: leased and
-fetched before history); the catalog API, Newznab (`/api`) and the web
+fetched before history; since 2026-10-27 a round grows the group's open
+tail range, a planned one at the tail's priority nothing spooled or
+took, up to RANGE_ARTICLES before it adds rows, as a tail nothing leased
+had become a sliver a group a round, 127k rows of a median 14k articles;
+`ranges:merge-tail` packs what such rounds left); the catalog API,
+Newznab (`/api`) and the web
 interface, from voyager 1 (the current build), with categories from the
 name; voyager 1's 29 learned garbage signatures (`rules:signatures`,
 `database/data/voyager1_signatures.tsv`).
@@ -139,7 +151,9 @@ dropping garbage; 28 TB of compressed overview traffic. Throughput goal:
    indexed groups in the article's Xref, the one with the lowest priority
    number; see the group list) → group articles into files → files
    into releases (`lib/release` rules) → drop garbage (random name +
-   dropped signature or tool) → set aside per-article random subjects and
+   dropped signature or tool), bare executables, dropped posters, and the
+   single-archive fakes (stub archives and nonsense pieces, below) → set
+   aside per-article random subjects and
    obfuscated subjects (counted only; phase 2). Returns per batch: release
    chunks, edge fragments, probe candidates, counters. Holds no state beyond
    its lease.
@@ -248,7 +262,10 @@ corpus filtered this way no longer serves a sweep of those two rules.
 Windows by posting time (2026-10-25; voyager 3's test data, LOG.md (104)):
 `ranges:dates` samples when each active group's articles were posted (one
 XOVER of 32 articles every 500M articles of the group, 16 … 1000 samples,
-the median Date of each slice, kept monotone; `group_dates`) and estimates
+the median Date of each slice, kept monotone; `group_dates`; over plain
+connections: Eweka cuts a compressed reply at the first ".\r\n" in its
+output, which some small replies begin with, `TruncatedReply`, LOG.md
+2026-10-25 (105)) and estimates
 every range's posting date from the curve (`ranges.est_date`; the edges of
 a window are off by up to one sample interval, three weeks in boneless).
 `ranges:capture --from 2025-01-01 --to 2026-01-01` marks every active
@@ -324,8 +341,13 @@ or music name whose content is a program: archive entries all
 executables, or the file itself one by name or by its first bytes, sniffed
 `exe`), *repeated* (the same first 16 KiB and size in 20+ releases),
 *locked* (passworded, no password from a spot, its NZB or the title),
-*unusable* (random names, no inner name, no par2 or inner-name merge, no
-spot, a day after settling). Verdicts (`release_verdicts`, with the rule
+*tiny* (one file of one article, 2–100 kB, under a readable release-like
+name outside the music, e-book and comic groups, without an e-book,
+subtitle, crack or sidecar word: the catalog's side of the edge's
+tiny-archive rule, 1.5M rows of the 2015–17 flood that the edge indexed
+before the rule; LOG.md (131), (135)), *unusable* (random names, no inner
+name, no par2 or inner-name merge, no spot, a day after settling).
+Verdicts (`release_verdicts`, with the rule
 set's version) hide releases from search (`judged` shows them); a verdict
 whose rule no longer matches is taken back. Nothing is dropped yet.
 
@@ -387,32 +409,82 @@ little in a backfill. That churn is most of the WAL (0.6 TB a day) and the
 releases, a third of them one poster's image dump (27.7 M rows, LOG.md
 (102)).
 
-1. **Blocks.** A group's history is still fetched as ranges of 2 M
-   articles in chunks of 100 k (the spool, the corpus, the prefetcher stay),
-   but it is *leased* as blocks: consecutive ranges covering at least 30
-   days of posting time (`ranges.posted_low/high`), so the window below is
-   small against a block (≤ 3% of a block's releases touch an edge). A
-   small group is one block; boneless has a block a week or so. Blocks of
-   one group run in parallel on different workers, in any order.
+1. **One cursor a group (amended 2026-10-26; was: blocks).** A group's
+   history is still fetched as ranges of 2 M articles in chunks of 100 k
+   (the spool, the corpus, the prefetcher stay) and leased as ranges, but
+   a group's ranges are indexed in article order by one worker at a time,
+   and the window below is *carried* from range to range: the next range
+   finishes what the previous one left open. The first design leased
+   blocks of 30 days of posting time, ran a group's blocks in parallel and
+   joined their edges afterwards (a block merger). The user asked why
+   (LOG.md 2026-10-25 (110)): downloading stays parallel through the
+   prefetcher and a Rust thread parses faster than a connection delivers,
+   so one worker a group loses nothing but the end of a reset, when the
+   last big group runs alone. The merger, the block cutting and the
+   `blocks` lease go. Costs: one seam a group where the history meets the
+   tail (the releases in flight then end as two incomplete halves; history
+   runs forward in one pass a group, priority only between groups), and
+   the carried state must survive a restart (5 below). Measured on six
+   consecutive ranges of a.b.ath (2016): one range is 15 hours of posting,
+   so without the carry no release would finish inside a range (LOG.md
+   2026-10-26 (111)).
+   **The group lease (2026-10-26, the user's question in LOG.md (117)):**
+   a worker thread leases a *group*, not a range, and asks the coordinator
+   for the group's next range each time it finished one; the coordinator
+   keeps one cursor a group (`group_leases`: the holder, its heartbeat,
+   the next range in article order, the snapshot it stands at). The
+   answer is the range (id, span, resume point), the rules version and the
+   stream URL — nothing about where the chunks are: the stream serves a
+   stored range's chunks from the corpus file and a live one's from the
+   spool, so a worker needs no mount and no corpus layout, and a remote
+   worker is the same binary with another URL. The overhead is a copy
+   through the node API: the corpus is ~30 B an article compressed (2016;
+   16–18 for 2021 and 2025), eight local workers at the bench's rate are
+   ~170 MB/s, which the HDD under the corpus bounds either way; a remote
+   worker is bounded by its link (100 Mbit ≈ 400k articles/s). History
+   therefore no longer waits for the corpus fill: a group's stored ranges
+   are read from disk, an unstored one is fetched live and captured into
+   the corpus on the way, and the prefetcher runs ahead of each held
+   group's cursor (plus the capture work left). A group with no planned
+   range left is the tail: the worker keeps the group and asks again
+   every minute, the planner's tail ranges fall into its order. A lease
+   whose holder went quiet for LEASE_TTL_SECONDS is taken over: the new
+   holder downloads the group's snapshot (5) and redoes the ranges since
+   it; the batch store's keys make the redo idempotent. The replay
+   harness (8) stays for offline runs; the driver on homelab goes once the
+   worker runs.
 
 2. **The window.** The worker reads a block in article order and keeps its
    open files and releases in memory (counters, names, sizes; the segment
    lists spooled to a file per block on the worker's disk). The *frontier*
-   is the median Date of the last 4,096 articles with a plausible date
-   (inside the block's span ± 7 days); Date headers alone aren't trusted
-   (clock skew, bogus years). A release is **finished** when the frontier
-   is 24 h past its last part (`WINDOW_HOURS`), or at the block's end:
-   written once, with its completeness decided then (every announced file
-   there; par2 decides later for those without a counter), never updated by
-   the header pass again. Incomplete releases are kept (user, 2026-10-24).
-   A fill posted after the window is a release of its own (the judge's
-   `repeated` rule already knows reposts). Releases still open at a block's
-   end are written flagged `edge`; the block merger (replaces the fragment
-   merger) joins the edge releases of adjacent blocks of a group by
-   identity key once both are done, and clears the flag.
-   The settler goes: `settled_at` is the finish. The **tail** keeps its
-   latency: a tail lease emits a release provisionally as today and the
-   window finalizes it once (two writes, not one per range).
+   is the median Date of the last 1,024 articles with a plausible date
+   (within 7 days of the frontier; 1,024 such dates in a row that agree
+   among themselves move it, the group having paused); Date headers alone
+   aren't trusted (clock skew, bogus years). 1,024, not more: the median
+   trails the newest article by half the ring, a day in a group of 500
+   articles a day. A release is **finished** when the frontier is 24 h past
+   its last part (`WINDOW_MS`): written once, with its completeness decided
+   then (every announced file there and no file short of parts; par2
+   decides later for those without a counter), never updated by the header
+   pass again. Incomplete releases are kept (user, 2026-10-24). A fill
+   posted after the window is a release of its own (the judge's `repeated`
+   rule already knows reposts). At a lease's end (the tail's poll, a worker
+   stopping; amended 2026-10-26, was: the block's end) the releases the
+   window hasn't closed on are written **provisionally** (`settled_at`
+   null, `finished: false`, flagged `edge`): a release being uploaded
+   shows in the catalog, and the carried window finishes the same row
+   later, since the release key is the identity, the group and the first
+   article; the store's finish path adds the files not sent before. No
+   merger: the carry does the joining. The settler goes: `settled_at` is
+   the finish. The **tail** needs no path of its own: its lease ends leave
+   such provisional releases, the next lease's continuation finishes them.
+   One exception (2026-10-27): a provisional release that is complete
+   (every file its counter announced found) and quiet for six hours is
+   settled by the coordinator (`Settler.settleComplete`, every five
+   minutes), as the finish would: no counted file can still come, and the
+   continuation may be hours away or never (the first run's range leases
+   left 109k provisional rows at their blocks' end edges, 606 of them
+   complete, shown as still collecting; the user, release 10697).
 
 3. **Albums.** Files of one poster in one owner group whose subjects differ
    only by a number sequence, or share a directory-like prefix, posted
@@ -420,11 +492,35 @@ releases, a third of them one poster's image dump (27.7 M rows, LOG.md
    shared template (or the prefix), its files kept by name so each stays
    findable; at least two files, else the single stays a release. This is
    for subjects that name no release (today's `post`/`stem` singles); a
-   poster's ten different cracks in a day stay ten releases.
+   poster's ten different cracks in a day stay ten releases. The stake,
+   from voyager 2's final analysis (LOG.md (112)): 77% of its 83.7 M
+   releases had one file, 50% one segment (0.3% of the segments);
+   a.b.pictures.erotica.porn alone was 27.7 M single-file releases of one
+   poster. A counter total shared by many stems of one poster within the
+   window is a batch counter, not a file count (Yenc-PP-A&A's [n/68067]):
+   `file_total` null for them, not 68,067 incomplete singles.
+   **Built 2026-10-26** (edge/src/indexer.rs, LOG.md (113)): a release the
+   window closed on with one file named by its stem or post, the name with
+   a number and no archive, volume, sidecar or random name, waits in the
+   album of its poster and template (the name's digit runs as `#`, the
+   extension apart: `#_#.jpg.data`) posted within WINDOW_MS of the album's
+   first file, at most ALBUM_MAX_FILES (1,000) a album; the album closes a
+   window after that span and is written as one release (`source: album`,
+   no file total) named by the members' shared numbers (`1481439079025_#
+   .jpg.data`), a lone member as the release it was. The batch store writes
+   the members' names as `release_texts` kind 6 (`member`, the first 20).
+   The directory-like prefix variant is not built (no case seen in the
+   corpus samples). The batch counter: a total ≥ 100 under which one poster
+   started 20+ releases within 100 batches gives none of them a file
+   total. On six a.b.ath ranges of 2016: 46 albums of 790 files among
+   2,099 chunks, most of them a subject parser miss (`[n/60] Shogun I
+   (1980) Shogun I (1980).avi.012`: the counter text taken into the name,
+   a volume each) that the album rule glues back together.
 
 4. **A slimmer catalog.** `releases` loses its text copies: `poster_id` →
    `posters(id)` (the table gets an id; 33 B → 4), `group_set_id` →
-   `group_sets` (the ~2,600 distinct group combinations; 85 B jsonb → 4),
+   `group_sets` (the distinct group combinations: 139,927 in voyager 2's
+   83.7 M releases, LOG.md (112); 85 B jsonb → 4),
    `release_key` and `identity_key` as 20-byte `bytea` (41 B each → 20),
    `source` a smallint. ~150 of ~430 B a row; the poster and key indexes
    shrink with them (~12 GB of 50 today); `releases_category_newest_idx`
@@ -437,7 +533,94 @@ releases, a third of them one poster's image dump (27.7 M rows, LOG.md
    a day; at a file per ~50 articles that is ~1 M open files of ~100 B in
    memory per block, and their segment lists (~25 B each) in the block's
    spool file, read back when a release is finished. Per worker thread a
-   few hundred MB at worst.
+   few hundred MB at worst. The carried window (amended 2026-10-26) has to
+   survive a restart (deploys restart the workers daily): its counters and
+   names are small (kilobytes for most groups, ~100 MB for the busiest),
+   the open files' segment lists are the bulk (~1.25 GB for a day of the
+   busiest group). A snapshot at every 2 M range would be ~600 B/article
+   of writes, so the state is snapshotted at a lease's end and on a timer,
+   to a file on bulk, and a crash redoes the ranges since the last snapshot
+   from the corpus or the prefetch cache: the deterministic keys make the
+   redo idempotent. **The snapshot goes through the API** (2026-10-26):
+   `PUT /node/groups/:id/snapshot` with the serialized window (the
+   frontier, the open files and releases, the open albums, the upload
+   numbers, the batch-counter table; gzip) and the article it stands at
+   (the last batch's `next`), kept by the coordinator on bulk as
+   `<snapshot dir>/<group id>`; `GET` gives it back to the next holder. A
+   worker snapshots when it gives a group back (a stop) and every
+   SNAPSHOT_MINUTES while it holds one, always right after a batch was
+   stored, so the snapshot's article is a resume point. The replay
+   harness (8) keeps a group's window in one process without one.
+   **Built 2026-10-26** (LOG.md (117)): migration 6500 `group_leases`;
+   `app/services/group_leases.ts` (lease by the most urgent planned
+   range within LEASE_MIN_PRIORITY..LEASE_MAX_PRIORITY, the longest free
+   first; takeover after LEASE_TTL_SECONDS, the ranges from the snapshot's
+   article on planned again; next; heartbeat; release; the snapshot
+   streamed to `SNAPSHOT_DIR/<group id>.gz`); the node API's
+   `groups/lease`, `groups/:id/next|heartbeat|release|snapshot`; the
+   range stream serving a stored range's chunks from the corpus; the
+   prefetcher ahead of held groups' cursors, skipping chunks the corpus
+   has; `RangeIndexer::for_group/begin_range/snapshot/restore` (bincode,
+   the files a table, SNAPSHOT_VERSION 1); `edge-node` holding groups by
+   default (`--ranges` keeps the old path). Tests: the lease order,
+   takeover and redo, the endpoints with a corpus-served stream and the
+   snapshot through the API, the snapshot restoring a window between two
+   ranges (the two-range release one, the album joined). Deployed and
+   started 2026-10-26 22:37 (LOG.md (118)). A chunk that doesn't inflate
+   (a live reply cut short; the spool doesn't check) is reported after
+   the batch before it went out and the range asked for again from that
+   chunk: the window stays (the first run lost it, and the open batch).
+   The lease bounds alone don't select a window: the history planner's
+   oldest buckets are priority 0–2 too, so the first run also took 2026
+   history at 0 (25k ranges in 40 groups, fetched live) while 1,739 of
+   the 2016 corpus ranges sat at voyager 2's leftover priorities (done
+   when the window was marked, planned again by the reset): hence
+   `LEASE_CAPTURE_ONLY` (the lease, next and the prefetcher's run-ahead
+   take capture ranges only) and the one-off repair of the priorities.
+   The stream polls the corpus as well as the spool while it waits for a
+   chunk: the corpus fill claims chunks of the same range ahead of the
+   leased run (the prefetcher's claims are per range and chunk), stores
+   them in the corpus only, and the run leaves them there; a stream that
+   looked at the corpus once, seconds before that store, waited its whole
+   hour on the spool and the worker lost the group and its open batch
+   (three groups on the first morning). A failed stream now posts the
+   open batch before the group goes back, like a stop does, so the
+   window's snapshot survives it; only a batch the coordinator refused
+   (lease lost, too large) still loses the window.
+   **Two lanes a group (2026-10-27; migration 6600, LOG.md (119)).** A
+   group's ranges in article order through one window meant a thread
+   that leased a group for its tail range first waded through the
+   group's history (older articles come first), and a held history
+   group's tail waited behind days of its ranges; and the bounds that ran
+   the 2016 window alone (0..0, capture only) shut the tail out entirely.
+   Now `group_leases` has a row per group and *lane*: the `history` lane
+   works the ranges the bounds hand out (the windows, the history), the
+   `tail` lane the tail planner's (priority at or below the tail's,
+   whatever the bounds), each with its own holder, window, snapshot
+   (`<group>-tail.gz`) and takeover redo (the lane's ranges only, and
+   only those the window did after the snapshot: leased, done since it,
+   or read past it). The edge puts `--tail-threads` of its threads on the
+   tail lane (EDGE_TAIL_THREADS; 1 of 2 or more by default, which is
+   what homelab's 16 run with), the rest on history; the prefetcher runs ahead of every held
+   cursor. The seam a group is the same one as before, once: where the
+   history lane reaches the first range the tail lane did, it is caught
+   up. **The lease order** within a lane and priority is a mix (the
+   user, 2026-10-27): one lease starts from the group with the most
+   planned articles, the next from the longest unheld one (never held
+   first), alternating on a sequence (`group_lease_turns`), so a big
+   group snapshotted at a stop doesn't wait behind hundreds of small
+   never-held groups and a quiet group's tail is visited all the same.
+   **`coordinator:status`** (`app/services/status.ts`, `--json`) prints
+   the catalog's room against `CATALOG_MAX_NVME_GB`, the ranges by band
+   (tail, raised, each capture window, the history by thousand), the
+   last hour's indexed and stored articles, the spool, the tail backlog,
+   the held groups with their snapshots' age and size, the snapshot
+   files, and the phase-2 backlogs (provisional releases, probe
+   candidates, the pending ladder, the judge's dirty marks). **End to
+   end** (`tests/e2e`, suite `e2e`, CI builds the binary first): the Rust
+   worker against the real node API, database, spool and corpus — the
+   corpus-fill race, a failed stream and the restart that restores the
+   snapshot, a restart whose snapshot file is gone, a takeover.
 
 6. **The corpus stays as it is.** Measured on 300 unfiltered chunks
    (30 M lines, LOG.md (102)): 43% of lines lack a "(k/N)" counter, and of
@@ -469,6 +652,237 @@ releases, a third of them one poster's image dump (27.7 M rows, LOG.md
    verdicts, file hashes. Resets are paused (2026-10-12); this one waits
    for the user's go-ahead, and for the new NVMe.
 
+### Schema (built 2026-10-25; migrations 6000 and 6100)
+
+Migration 6000 *is* the reset: it truncates every table the indexing
+recomputes and refuses to run on a catalog with releases unless
+`VOYAGER3_RESET=1`, so merging `voyager-3` into `rebuild` is the go-ahead.
+Survives: `groups`, `ranges` (planned again from their start, with their
+priorities, `capture`, `est_date`, `stored_at`, `plain_chunks` and measured
+`posted_low/high`), the corpus, `probe_cache`, `spots` (unlinked from
+releases; the inspector matches them again), `nntp_shares`, `nodes`,
+`users`, the rule sets.
+
+- `releases`: created anew, 8-byte columns first: `id`, `posted_at`,
+  `last_posted_at`, `first_article`, `last_article`, `segments`, `bytes`,
+  `settled_at` (the finish; null while a tail release is provisional),
+  `inspected_at`, `first_seen_at`; `group_id`, `group_set_id`, `poster_id`,
+  `file_total`, `files`, `chunks`, `category`; `source smallint`
+  (`lib/release/source.ts`: 1 subject, 2 post, 3 stem, 4 spot, 5 nzb, 6
+  album); `readable`, `has_par2`, `complete`, `passworded`, `sidecar_only`,
+  `edge` (open at its block's end); `release_key` and `identity_key` as
+  20-byte `bytea`; `name`, `inner_name`. Gone: `poster`, `groups`,
+  `last_seen_at`. Indexes: pkey, release_key unique, (identity_key,
+  group_id), newest, (poster_id, newest), passworded, `(group_id,
+  identity_key) WHERE edge` for the block merger, `id WHERE settled_at IS
+  NULL` for the provisional ones. No trigram indexes: search matches
+  `release_texts`. The category trigger reads the groups from
+  `group_sets`; the judge's dirty trigger watches `poster_id` instead of
+  `poster`.
+- `pending_releases` (migrations 6300 and 6400, the HDD): the unnamed
+  releases outside the catalog (decision f): `release_key` (primary),
+  `identity_key`, `group_id`, `name` (the random stem), `poster` text, the
+  posting span, `first_article`, `last_article`, counts, `file_total`,
+  `partial`, `finished`, `tool`, `candidates` jsonb (the probe targets),
+  `inspections` jsonb (what each probe showed), `named`, `status` (0
+  waiting, 1 named and in the catalog, 2 discarded, 3 exhausted, 4 named
+  awaiting the move, 9 taken), `tries`, `created_at`, `probed_at`; partial
+  indexes on the waiting (by `created_at`) and the named (by `probed_at`).
+- `pending_signatures` (6400): per (tool, group, poster; poster '' for the
+  tool and group together) `probed`, `named`, `dropped`, `updated_at`.
+- `release_texts(release_id, kind smallint, text varchar(255))` (6200),
+  primary key on all three, one trigram index (fastupdate): every string
+  search matches against, appended by whoever learns one and never an
+  update of the releases row. Kinds (`lib/release/text_kind.ts`): 1 name
+  (the release name, stored once more so one index serves everything),
+  2 file (the prober's names inside an archive), 3 nfo (an NFO's title),
+  4 par2 (a set's name), 5 media (a container's title); at most 20 per
+  release and kind (`textsToStore`). A repeat insert is a no-op; the
+  judge's dirty mark follows an insert or delete, so a rule can react to a
+  name arriving later. Search: the trigram index yields release ids, joined
+  to `releases` by primary key (never an OR across the two tables, which
+  the planner cannot serve from indexes). `inner_name` stays on `releases`
+  as the one display string. Facets to filter on (resolution, codec,
+  language) are not texts: columns, or a `meta jsonb` with a
+  `jsonb_path_ops` index once one is proven useful.
+- `posters(id serial, poster unique)` with its trigram index;
+  `group_sets(id serial, groups jsonb unique)`, a sorted array of names.
+- `blocks` (6100): `group_id`, `start_id`, `end_id`, `next_id` (the resume
+  point), `posted_low/high`, `status`, `priority`, `node_id`, `attempts`,
+  `rule_version`, `headers`, `releases`, `leased_at`, `heartbeat_at`,
+  `done_at`, `error`; indexes todo (priority, id) and leased.
+  `ranges.block_id` points at the block; the ranges' own lease columns stay
+  until the leases move (then a migration drops them).
+- `release_chunks`, `rule_counts`, `rule_samples`: `range_id` → `block_id`.
+- Gone: `fragments`, `range_blocks`, `release_final()` (final = finished:
+  `settled_at IS NOT NULL`), the settler, the fragment merger,
+  `releases_category_newest_idx`.
+- `release_inspections` and `probe_candidates` were on `CHUNK_TABLESPACE`
+  (the HDD) with their indexes; both are back on the NVMe (migrations 4800
+  and 7300). Note LOG.md 2026-10-19 (81): the judge's fake rule read all of
+  `release_inspections` every round, 17 min on the HDD; and 2026-11-03
+  (136): the prober's catalog claim updates `probe_candidates` from random
+  points every minute, 2–3 min on the HDD beside the pending ladder's sweep.
+  What a worker reads or updates by random key every minute stays on the NVMe.
+
+### Batch store (built 2026-10-25)
+
+`app/services/batch_store.ts`: every release chunk of a batch is one
+release, inserted once. Its keys (`lib/release/keys.ts`): `identity_key`
+the identity's SHA-1 as 20 bytes; `release_key` the SHA-1 of the identity,
+the owner group and the release's first article, so it is deterministic
+(no numbering under a group lock, no catalog lookups before writing) and a
+repost, starting at another article, is a release of its own. Its poster
+and group set by id (`app/services/catalog_refs.ts`: insert what's new,
+read the ids, cache them per process; the tests forget the cache per
+test), its source as a code, its name into `release_texts` (kind 1). The
+chunk's own counts go in as the release's; `complete` is decided then from
+the counter. A finished release's probe candidates are claimable at once
+(`ready`, `unreadable`); a provisional tail release (`finished: false` on
+the chunk, `settled_at` null) is finished by a later chunk of the same
+release key, which adds the files not sent before (sums, one more chunk),
+sets `settled_at` and makes its candidates claimable; a finished release
+sent again is left as it is (`ON CONFLICT … WHERE settled_at IS NULL`).
+Gone: the join of a chunk into an earlier release of its key (the window
+keeps an upload together; the block merger joins what a block's end cut),
+the join across groups (FTD's spread uploads: a gap, to be handled by the
+block merger or the par2 merger across groups), the numbering of release
+keys, the group lock, `fragments` (ignored when sent; the Rust edge still
+sends them). The chunk type carries `finished` and `edge` (both optional;
+the Rust edge doesn't send them yet, so its releases are finished, not
+edges). Interim until blocks are leased: chunks, rule counts and samples
+are keyed by `ranges.block_id`, and a range without a block gets a block
+of its own in the store (`blockOf`).
+
+Readers moved with it: search joins `posters` and `group_sets`, matches
+words in `release_texts` (`r.id IN (SELECT release_id … ILIKE)`), searches
+posters by id, filters `source` by code and returns the source name; the
+par2 merger, the sidecar folder and the judge's scraped rule use
+`poster_id` (`release_merges.from_key` holds the key as hex); the spot
+inspector matches by bytea identity keys and inserts its spot releases by
+id references with their title in `release_texts`; the spotter and the NZB
+service compare the source code. Tests insert fixtures through
+`tests/functional/helpers/releases.ts` (`insertRelease`: text keys, poster
+and groups by name, source by name). The unit and functional suites pass.
+
+### The window in the Rust edge (built 2026-10-25)
+
+`edge/src/indexer.rs`: the scanner and the file aggregator as before; then
+the window instead of the range path. `Frontier` keeps a ring of the last
+`FRONTIER_ARTICLES` (1,024) plausible dates (within `PLAUSIBLE_MS`, 7
+days, of the frontier; 1,024 implausible dates in a row that agree move
+it), recomputed once per chunk; established after 64 dated articles
+(`settle` at the block's end for smaller blocks); its first value is the
+block's start in posting time. Per chunk: completed files, then the files
+whose last part is before frontier − `WINDOW_MS` (24 h; undated ones idle
+for `idle` chunks), go into the open release of their key whose posting
+span is within `UPLOAD_GAP_MS` (3 h) of theirs, else start one (the
+upload number per key, distinct within a batch); releases whose last part
+is before frontier − `WINDOW_MS` are judged and written `finished`, as an
+`edge` when posted within `WINDOW_MS` of the block's start. At the end
+(`finish`) everything left is written: finished when the window closed on
+it, else `finished: false, edge: true`. A file short of parts stays in
+its release and sets `partial` (the store: `complete = false`). A lone
+article of a multi-part random file is still counted `per-article`, not
+written; before the block's end a scrambled name counts as random too
+(`looks_scrambled`: a random subject and poster per article, 160k of
+alt.binaries.storage's releases; LOG.md (122)).
+
+**Single-archive rules (built 2026-10-31; LOG.md (123)-(127)).** The judge
+(`names::judge`, `RuleSet.judge`) gets each file's articles seen and their
+bytes, and whether the release is *settled* (the window closed on it: a
+par2 can still join an open one; the start edge doesn't count, since a par2
+follows its rar and 7% of the catalog's releases are start edges). A settled
+release of one unsplit zip, rar or 7z and nothing else is dropped as a
+**stub archive** (`stub-archive`, key: the legs tripped) when posted to
+2+ groups, under 100 MB (the fakes stop at ~90 MB; hash-named uploads,
+real albums and passworded videos above it don't), and (A) its articles
+are ≤ 256 KiB with the release ≥ 1 MB (Power-Post's 250 kB default; a
+one-article release always has small articles, which exempts e-books and
+sidecars), or (X) posted to 4+ groups, or (W) by a webmail poster. A
+ROT13 volume name (".cnegN.rar" is ".partN.rar": a.b.erotica's passworded
+videos, LOG.md (128)) is a volume, not a single archive. A release
+touching a music, e-book, comics, tabs or fonts group is exempt: those
+communities post real single small archives the same way (karaoke zips to
+a.b.sounds.karaoke + a.b.test with Power-Post's articles; 16 of 32,308
+known fakes lost, LOG.md (127)). These are the
+exe-stub and recycled-installer floods under film and software names:
+55% of the catalog on 2026-10-30, 99.3% of the known fakes, 0 known
+content in the 217k sample, verified on 1,500 decoded articles. A settled
+single archive of one article and 100 kB–1 MB under a release-like,
+unscrambled name without spaces and without an e-book format in it is
+dropped as a **nonsense piece** (`nonsense-piece`, key: the upload tool):
+the newznzb.* pieces (scene-styled or random with a "-XX" suffix), 9% of
+the catalog; under 100 kB are cue/log sidecars and adverts, and the names
+with spaces are books and albums. Below that, a settled single archive of
+one article and 2–100 kB under a release-like name (spaces allowed), outside
+the exempt groups and without an e-book, subtitle, crack/keygen/trainer,
+cover or sidecar word, is dropped as a **tiny archive** (`tiny-archive`,
+key: the upload tool; LOG.md (131), (133)): the 2015–17 flood's RARs
+holding "<title>.exe" at 2–60 kB, 1.66M catalog rows from a dozen posters
+(12 of 12 fetched fake); the word exemptions keep the real tiny archives
+(subtitles, cracks, trainers) at the price of ~50k fakes named "Incl
+Keygen" that the prober judges later. The rules are switched by the
+rule set (`dropStubArchives`, `dropNonsensePieces`, `dropTinyArchives`),
+sampled like every drop, and mirrored in `lib/rules/rule_set.ts` (the TypeScript indexer
+judges as each batch completes files, so it passes settled without the
+window's knowledge). Gone: fragments and the cut-off, `uploads_of` (online now),
+`check.ts` and the fixture comparison (the parity with the TypeScript
+indexer; CI runs `edge/bench-fixtures.sh` instead), the `incomplete` count
+(incomplete releases are kept). The wire carries `finished`, `edge` and
+`partial`; the store reads them (`partial` into `complete`). Memory: the
+window holds a day of a group's segments in memory (the busiest group's
+~50 M); the spool file per block of the design is not built yet.
+
+### The replay harness (built 2026-10-26)
+
+`node ace replay:corpus --group <name> [--dir CORPUS_DIR] [--from article]
+[--to article] [--limit ranges] [--batch-chunks 5] [--no-finish]` (commands/
+replay_corpus.ts; design 8): a group's complete corpus ranges in article
+order through one window in the Rust edge (`edge-bench --group <dir>
+--rules rules.json --wire -`: the window carried from range to range, the
+end finished after the last range; `edge/src/lease.rs` has the lease JSON
+the worker and the bench share) and the batch store into a scratch
+catalog, as the worker would store them: each range leased to a node named
+`replay`, its batches stored at its resume point (pulled from the bench's
+stdout as the store keeps up), the range finished when its last batch is
+in. It refuses a database not named `*_replay` or `*_scratch`.
+The rules are the scratch catalog's current set and group priorities:
+copy `groups` and `rule_sets` from the catalog for a faithful replay. The
+report: releases finished, provisional, edges; complete, incomplete,
+undecided; readable, with par2; files, segments as a share of the
+headers; bytes per header of the NVMe tables (releases, release_texts,
+posters, group_sets, rule counts and samples) and of those on
+CHUNK_TABLESPACE (release_chunks, probe_candidates, release_inspections).
+The app image carries `edge-bench` for it (Dockerfile); on homelab the
+scratch catalog is a second database (`usenet_replay`, default tablespace
+`nvme` like the catalog, the chunk tables on `pg_default`; migrated with
+the voyager-3 image built from the deploy cache's repo, `groups` and
+`rule_sets` copied over), the corpus mounted read-only.
+`deploy/replay-window.sh` runs a window there: each group whose window
+ranges are all stored is replayed once with the window's bounds as
+`--from`/`--to` (the corpus holds other windows of the same group), 4
+groups at a time, logs and `.done` markers in `.cache/replay/`, looping
+until the window is done (LOG.md 2026-10-26 (111) has the first local
+numbers, (112) the homelab run). **Deployed 2026-10-26** (LOG.md (116)):
+voyager 3 runs on homelab without its edge worker (`.cache/no-edge`:
+deploy.sh and the nightly resume leave it stopped, since range by range
+it would split releases at every edge with no merger left), and history
+comes from the corpus through this driver into the catalog itself
+(`DB=usenet_rebuild OUT=.cache/replay-catalog`, the harness allowed by
+`REPLAY_INTO_CATALOG=1`), the prober's ladder, the merger and the judge
+working on it as it fills; the scratch run's numbers (63 groups) were
+read before (LOG.md (115)). The tail waits for the window worker.
+
+Next: the replay on homelab over the 2016 window (the NVMe bytes per
+header for real), the window worker (design 1 and 5: group leases, the
+stream serving stored chunks, the snapshot through the API; being built,
+LOG.md (117)), the prober writing `release_texts` kind 2, the spot/NZB
+match on pending keys (decision f), the tail's provisional writes at a
+range end (the window's open releases shown while they upload: the files
+sent provisionally marked, the finish sending the rest). Gone from the list: block
+leases, the planner and the block merger (design 1, amended).
+
 ## Decisions
 
 a. Chunk payload: a write-once `bytea` table in Postgres, partitioned by
@@ -478,11 +892,84 @@ c. Group list: editable, owner by priority number (above). **Accepted.**
 d. Reposts: separate releases in phase 1, merged in phase 2. **Accepted.**
 e. Headers through the coordinator's relay, with prefetch (part 1).
    **Accepted.**
-f. Random-named releases (no file with a readable name) keep no chunk in
-   the header pass: their file keys and probe candidates only, until
-   phase 2 names them (bodies) and fetches their segments again. The full
-   history is ~3 TB with them, ~1.2–1.4 TB without (725 M of the 1.25 B
-   kept files; LOG.md 2026-10-09). **Accepted** 2026-10-09, not built yet.
+f. **Nothing enters the catalog without a usable name** (amended
+   2026-10-26, the user's direction; before: random-named releases keep
+   no chunk but their rows). A release the window closed on whose files
+   all have random names (`isUnreadableName`, `names::is_unreadable_name`:
+   a name read token by token, so separators, spaces and random
+   "extensions" don't hide a scramble; since 2026-11-02, LOG.md (129),
+   when 15% of the catalog's "readable" names were "Zx5AjNaK-mh7CFg6Hf8uu"
+   and "Description - Lt5nHjlo3RiuBb3C") is no release yet: it goes to `pending_releases`
+   (migration 6300, the HDD) with its keys, counts, article span, upload
+   tool and probe targets, no row in `releases`, `release_texts` or
+   `posters`, no chunk. The pending ladder works them **from least to most
+   effort**: (0) no fetch: a signature verdict (tool and group, poster)
+   learned from earlier probes, a spot or NZB that names the message-ids;
+   (1) a par2 main file's first article (it names the files); (2) the first
+   article of the first file (the yEnc name, an archive's listing in its
+   first 16 KiB); (3) a 7z's tail; then nothing more. A probe that yields
+   a usable name moves the release into the catalog, its segments read
+   again from the corpus (the range stored) or the server; one that
+   yields none counts against its signature, and a signature with
+   ≥100 probes and ≤0.5% named (voyager 2's garbage learner, LOG.md
+   2026-10-08 (5)) is **discarded** from then on, a sample of it still
+   probed to notice a change. **By class** (2026-11-02, LOG.md (130)): a
+   pending release's class is its tool, group, size bucket, name shape and
+   partial flag (`lib/release/classes.ts`, `names::pending_class`, the SQL
+   `pending_class()`; `pending_releases.klass`), and `pending_classes` keeps
+   each class's climbs and names; the claim takes the classes with the best
+   record first (`classScore`, a prior for the unexplored, at most a quarter
+   of a claim from one class), the cheapest target article first within one
+   (`target_bytes`), and the oldest of the rest; a class with the learner's
+   thresholds is dropped like a signature, and the rule set's
+   `droppedClasses` keep its releases out of the pending store at the edge,
+   the explored sample excepted. A probed file whose first 16 KiB and size
+   match a named catalog release's (`file_hashes`) takes that name without
+   another fetch (`app/services/hash_twins.ts`), on the ladder and for
+   catalog releases without one. The reason, in the 2016 replay: one poster's
+   hash-named pieces ("Bob", 11 aliases, 3.2 M releases in five small
+   groups) were 76% of the rows and three quarters of the NVMe bytes for
+   11% of the segments and nothing anyone can find; the full history has
+   725 M random-named files (LOG.md 2026-10-09). Built 2026-10-26: the
+   edge sends an unnamed release with an empty payload and its tool, the
+   batch store writes `pending_releases` (LOG.md (114)); the ladder in the
+   prober (`app/services/pending_ladder.ts`, LOG.md (115)): a claim takes
+   the oldest waiting releases' next target (par2, archive, lone, 7z-tail
+   in that order), at least PENDING_SHARE (25%) of each claim next to the
+   catalog's own candidates, answered from `probe_cache` when the article
+   was read before; `usableName` (lib/probe/inspect.ts) takes the largest
+   named file of a par2 set or a listing, or a lone file's yEnc name, that
+   is neither packaging nor random; a name sets status 4, no name the next
+   target or status 3 (exhausted); each finished climb counts on
+   `pending_signatures` for (tool, group, poster) and (tool, group), and a
+   signature with PENDING_MIN_PROBES (100) climbs and at most
+   PENDING_MAX_NAMED (0.5%) named is dropped: a sweep beside the lanes
+   (every five minutes, in its own loop) discards the waiting releases
+   of each dropped signature and class once (`swept_at`; the edge drops
+   their new releases itself), the rules' `samplePercent` of them left to
+   climb, each by an index (`pending_releases_signature_idx` for a
+   poster's rows: tested on the rows, a signature was a scan of the whole
+   table; LOG.md (133), (134)). The batch store, the record, the claim
+   and the sweep lock pending rows in one order, by key and the counters
+   after (`lockOrder` for the multi-row statements; the record and the
+   sweep lock their rows with `SELECT … ORDER BY release_key FOR UPDATE`
+   first, since an UPDATE's join or `= ANY(subquery)` locks in plan
+   order), and a flush that deadlocks anyway keeps its outcomes for the
+   next (LOG.md (134), (137)). The namer
+   (`app/services/pending_namer.ts`, in the prober's process): a named
+   release's article span read again from the corpus (the stored range's
+   chunks) or by XOVER on a connection of its own, a chunk of articles
+   at a time (one string of a wide span topped V8's limit; spans over 3M
+   articles are given up), through the TypeScript
+   indexer with `partialFiles` (a file given up short of parts stays a
+   partial file of its release, as the Rust window writes it; without it
+   the range path's fragment rule left nothing to find: by 2026-10-29
+   about 52k of 210k named releases failed, all partial), written as a release under the pending key with `source`
+   weak-stem, the name as `inner_name`, its chunk and its candidates (the
+   prober answers those from `probe_cache`: inspections, par2 sets, hashes,
+   the password flag). Not built: the spot/NZB match on pending keys (step
+   0 of the ladder beyond signatures), and a signature by poster alone
+   across groups.
 g. The account's NNTP connections shared by priority (`nntp_shares`): the
    coordinator (headers) before the prober, each between its minimum and
    maximum, 5 kept for the spotter, the spot inspector, the plain

adonisrc.ts

@@ -91,6 +91,12 @@ export default defineConfig({
         name: 'functional',
         timeout: 30000,
       },
+      // The Rust edge worker against the real node API (tests/e2e): needs edge/target/release/edge-node.
+      {
+        files: ['tests/e2e/**/*.spec.{ts,js}'],
+        name: 'e2e',
+        timeout: 120000,
+      },
     ],
     forceExit: false,
   },

app/controllers/node_controller.ts

@@ -1,17 +1,15 @@
 import type { HttpContext } from '@adonisjs/core/http'
 import { Readable } from 'node:stream'
 import { setTimeout as sleep } from 'node:timers/promises'
+import db from '@adonisjs/lucid/services/db'
 import Leases, { LeaseLost } from '#services/leases'
+import GroupLeases, { LANES } from '#services/group_leases'
 import BatchStore, { StaleBatch } from '#services/batch_store'
-import Spool from '#services/spool'
-import { KEEPALIVE, encodeFrame, fromWire, type WireBatch } from '#lib/edge/index'
+import Spool, { type ChunkSpan } from '#services/spool'
+import Corpus from '#services/corpus'
+import coordinatorConfig from '#config/coordinator'
+import { KEEPALIVE, encodeFrame, fromWire, type Lane, type WireBatch } from '#lib/edge/index'
 
-/**
- * How long the stream waits for a chunk the prefetcher hasn't spooled yet. Long: a chunk that keeps
- * coming in isn't given up (only a stalled one is), and some groups' chunks take 20+ minutes
- * (a.b.0day.stuffz, 17 MB in 22 min; LOG.md 2026-10-10 (3)).
- */
-const CHUNK_WAIT_MS = 60 * 60 * 1000
 /** While waiting, an empty frame this often: an HTTP client gives up on a quiet body (Node's after 5 minutes). */
 const KEEPALIVE_MS = 20_000
 
@@ -30,18 +28,109 @@ export default class NodeController {
     return lease
   }
 
-  /** The range's chunks from its resume point, waiting for those still being fetched. */
+  // Group leases (REBUILD.md, "Voyager 3: window indexing", 1 and 5): a worker thread holds a group and takes its
+  // ranges in article order; the window's snapshot goes through here too, so a worker needs no disk of its own.
+  // `?lane=tail` is the group's tail cursor (its own holder, window and snapshot); history by default.
+
+  /** The lane the request names (`?lane=`), or null after a 400 for one that isn't a lane. */
+  private lane(ctx: HttpContext): Lane | null {
+    const lane = ctx.request.qs().lane ?? 'history'
+    if (LANES.includes(lane)) return lane
+    ctx.response.badRequest({ error: `lane: ${LANES.join(' or ')}, not "${lane}"` })
+    return null
+  }
+
+  async leaseGroup(ctx: HttpContext) {
+    const lane = this.lane(ctx)
+    if (!lane) return
+    const lease = await GroupLeases.lease(this.node(ctx), lane)
+    if (!lease) return ctx.response.noContent()
+    return lease
+  }
+
+  async nextRange(ctx: HttpContext) {
+    const lane = this.lane(ctx)
+    if (!lane) return
+    try {
+      const next = await GroupLeases.next(Number(ctx.params.id), this.node(ctx), lane)
+      if (!next) return ctx.response.noContent()
+      return next
+    } catch (error) {
+      if (error instanceof LeaseLost) return ctx.response.conflict({ error: error.message })
+      throw error
+    }
+  }
+
+  async groupHeartbeat(ctx: HttpContext) {
+    const lane = this.lane(ctx)
+    if (!lane) return
+    return this.lifecycle(ctx, (groupId, nodeId) => GroupLeases.heartbeat(groupId, nodeId, lane))
+  }
+
+  async releaseGroup(ctx: HttpContext) {
+    const lane = this.lane(ctx)
+    if (!lane) return
+    return this.lifecycle(ctx, (groupId, nodeId) => GroupLeases.release(groupId, nodeId, lane))
+  }
+
+  /**
+   * PUT /groups/:id/snapshot?next=N: the window's serialized state standing at article N, streamed to disk as sent
+   * (gzip; the body parser leaves an octet-stream alone, so the request is still readable here).
+   */
+  async putSnapshot(ctx: HttpContext) {
+    const lane = this.lane(ctx)
+    if (!lane) return
+    const next = Number(ctx.request.qs().next)
+    if (!Number.isFinite(next) || next <= 0) return ctx.response.badRequest({ error: 'next (the snapshot\'s article) is needed' })
+    const raw = ctx.request.raw()
+    const body = raw !== null ? Readable.from([Buffer.from(raw, 'binary')]) : ctx.request.request
+    return this.lifecycle(ctx, (groupId, nodeId) => GroupLeases.putSnapshot(groupId, nodeId, next, body, lane))
+  }
+
+  async getSnapshot(ctx: HttpContext) {
+    const lane = this.lane(ctx)
+    if (!lane) return
+    try {
+      const snapshot = await GroupLeases.getSnapshot(Number(ctx.params.id), this.node(ctx), lane)
+      if (!snapshot) return ctx.response.notFound({ error: 'no snapshot' })
+      ctx.response.header('Content-Type', 'application/octet-stream')
+      ctx.response.header('Content-Length', String(snapshot.size))
+      ctx.response.header('X-Snapshot-Next', String(snapshot.next))
+      return ctx.response.stream(snapshot.stream)
+    } catch (error) {
+      if (error instanceof LeaseLost) return ctx.response.conflict({ error: error.message })
+      throw error
+    }
+  }
+
+  /**
+   * The range's chunks from its resume point: a stored range's from the corpus (a worker needs no mount; REBUILD.md
+   * design 1), else from the spool, waiting for those still being fetched.
+   */
   async stream(ctx: HttpContext) {
     const nodeId = this.node(ctx)
     const rangeId = Number(ctx.params.id)
     const range = await this.leased(ctx, rangeId, nodeId)
     if (!range) return
     const spans = Spool.chunksOf({ start: Number(range.start_id), end: Number(range.end_id) }, Number(range.next_id))
+    const group = range.capture ? await db.from('groups').where('id', range.group_id).select('name').first() : null
+    const corpusRange = group ? { group: group.name as string, start: Number(range.start_id), end: Number(range.end_id) } : null
+    // The chunk from the corpus or the spool, whichever has it. Both are polled while waiting: the corpus fill
+    // claims a chunk ahead of the leased run (the prefetcher's claims are per range and chunk), stores it in the
+    // corpus only, and the run then leaves it there — a stream that looked at the corpus once, before that
+    // store, waited the whole hour on the spool (three groups on 2026-10-27, LOG.md (118)).
+    const stored = async (span: ChunkSpan) =>
+      corpusRange && (await Corpus.has(corpusRange, span)) ? Corpus.read(corpusRange, span) : null
     async function* frames() {
       for (const span of spans) {
-        const waitUntil = Date.now() + CHUNK_WAIT_MS
+        const waitUntil = Date.now() + coordinatorConfig.streamChunkWaitMs
         let quietSince = Date.now()
-        while (!(await Spool.has(rangeId, span))) {
+        let data = await stored(span)
+        while (!data) {
+          if (await Spool.has(rangeId, span)) {
+            data = await Spool.read(rangeId, span)
+            break
+          }
           if (Date.now() > waitUntil) throw new Error(`Chunk ${span.from}–${span.to} of range ${rangeId} wasn't spooled in time`)
           if (Date.now() - quietSince >= KEEPALIVE_MS) {
             quietSince = Date.now()
@@ -49,8 +138,9 @@ export default class NodeController {
             await Leases.heartbeat(rangeId, nodeId)
           }
           await sleep(500)
+          data = await stored(span)
         }
-        yield encodeFrame({ ...span, data: await Spool.read(rangeId, span) })
+        yield encodeFrame({ ...span, data })
         // Streaming keeps the lease alive.
         await Leases.heartbeat(rangeId, nodeId)
       }
@@ -75,7 +165,11 @@ export default class NodeController {
   }
 
   async finish(ctx: HttpContext) {
-    return this.lifecycle(ctx, (rangeId, nodeId) => Leases.finish(rangeId, nodeId))
+    return this.lifecycle(ctx, async (rangeId, nodeId) => {
+      const range = await Leases.leased(rangeId, nodeId)
+      await Leases.finish(rangeId, nodeId)
+      await GroupLeases.rangeDone(Number(range.group_id), Number(range.headers), GroupLeases.laneOf(Number(range.priority)))
+    })
   }
 
   /** POST /ranges/:id/corrupt { from }: a chunk that doesn't inflate; the range goes back. */

app/services/batch_store.ts

@@ -1,14 +1,20 @@
 import db from '@adonisjs/lucid/services/db'
 import { pruneSevenZipTails } from '#services/seven_zip_tails'
 import type { TransactionClientContract } from '@adonisjs/lucid/types/database'
-import { UPLOAD_GAP_MS, type EdgeBatch, type ProbeCandidate, type ReleaseChunk } from '#lib/edge/index'
+import type { EdgeBatch, ProbeCandidate, ReleaseChunk } from '#lib/edge/index'
 import ChunkPartitions from '#services/chunk_partitions'
+import CatalogRefs, { groupSetKey, posterText } from '#services/catalog_refs'
 import { decodeChunk } from '#lib/chunks/index'
 import { SIDECAR } from '#services/sidecar_folder'
+import { identityKey, releaseKey } from '#lib/release/keys'
+import { sourceCode } from '#lib/release/source'
+import { uploadTool } from '#lib/rules/index'
+import { byEffort, lockOrder } from '#services/pending_ladder'
+import { TEXT_KINDS, TEXT_MAX, textsToStore } from '#lib/release/text_kind'
 
 /**
  * A small new release (at most 3 files, under 1 MB) whose files are all sidecars (.nfo, .sfv, .nzb, …): the sidecar
- * fold's candidates (releases.sidecar_only; the settler queues only these). Its file list decoded here, small as it is.
+ * fold's candidates (releases.sidecar_only). Its file list decoded here, small as it is.
  */
 function sidecarOnly(r: ReleaseChunk): boolean | null {
   if (r.files > 3 || r.bytes >= 1_000_000) return null
@@ -19,18 +25,34 @@ function sidecarOnly(r: ReleaseChunk): boolean | null {
 export class StaleBatch extends Error {}
 
 export interface StoredBatch {
+  /** Release chunks in the batch. */
   releases: number
+  /** Releases the batch started. */
   newReleases: number
+  /** Provisional (tail) releases the batch finished. */
+  finalized: number
   candidates: number
-  fragments: number
+  /** Unnamed releases written to the pending store instead of the catalog (decision f). */
+  pending: number
 }
 
+/** A chunk's slot in its batch: the key and the upload the edge numbered it with (ProbeCandidate points at it). */
+const slot = (key: string, upload: number) => `${key}\n${upload}`
+
 /**
- * Stores one batch of a leased range (REBUILD.md, parts 3–7), in one
- * transaction: release aggregates in the catalog, one chunk per release,
- * probe candidates, rule counts and samples, edge fragments, and the
- * range's resume point. A batch that doesn't start at the resume point was
+ * Stores one batch of a leased range (REBUILD.md, "Voyager 3: window indexing", 2 and 4) in one transaction: each
+ * release chunk is a whole release the edge's window finished, written once (its poster and group set by id, its
+ * keys as bytea, its name into release_texts), with one chunk of its files, its probe candidates, and the batch's
+ * rule counts and samples; then the range's resume point moves. A batch that doesn't start at the resume point was
  * sent twice or too late, and is refused (StaleBatch), so the sums stay right.
+ *
+ * A tail release comes twice: provisionally (finished: false, settled_at null, its candidates waiting) and then
+ * finished, under the same release key (the identity, the group and its first article: lib/release/keys.ts), with
+ * the files not sent before (a chunk's counts are its own; the finish adds them up); the finish is the one update a
+ * release gets from the header pass. A finished release sent again is left as it is.
+ *
+ * Interim, until blocks are leased (the next step): chunks, counts and samples are keyed by the range's block
+ * (ranges.block_id), and a range without one gets a block of its own here.
  */
 class BatchStore {
   async store(rangeId: number, nodeId: number, batch: EdgeBatch): Promise<StoredBatch> {
@@ -39,7 +61,9 @@ class BatchStore {
       // One batch of a range at a time, without holding the range's row for the whole transaction: the
       // prefetcher's and heartbeats' updates of it waited on it (0.5 s on average, NNTP connections idle;
       // LOG.md 2026-10-09 (13)). The resume point is checked again when it moves, at the end.
-      await trx.rawQuery(`SELECT pg_advisory_xact_lock(hashtextextended('range:' || ?::text, 0))`, [rangeId])
+      await trx.rawQuery(`SELECT pg_advisory_xact_lock(hashtextextended('range:' || ?::text, 0))`, [
+        rangeId,
+      ])
       const range = await trx.from('ranges').where('id', rangeId).first()
       if (!range || range.status !== 'leased' || Number(range.node_id) !== nodeId)
         throw new StaleBatch(`Range ${rangeId} is not leased to this node`)
@@ -47,35 +71,48 @@ class BatchStore {
         throw new StaleBatch(
           `Range ${rangeId} is at ${range.next_id}, the batch starts at ${batch.from}`
         )
+      const blockId = await this.blockOf(trx, range)
       const result: StoredBatch = {
         releases: batch.releases.length,
         newReleases: 0,
+        finalized: 0,
         candidates: 0,
-        fragments: batch.fragments.length,
+        pending: 0,
       }
 
+      // Nothing enters the catalog without a usable name (decision f): an unnamed release (no readable file name)
+      // goes to the pending store with its probe targets, the named ones to the catalog.
+      const named = batch.releases.filter((r) => r.readable)
+      const unnamed = batch.releases.filter((r) => !r.readable)
       const written = await this.writeReleases(
         trx,
-        rangeId,
+        { groupId: Number(range.group_id), blockId },
         batch.from,
-        batch.releases,
+        named,
         batch.candidates
       )
       result.newReleases = written.newReleases
+      result.finalized = written.finalized
       result.candidates = written.candidates
+      result.pending = await this.writePending(
+        trx,
+        Number(range.group_id),
+        unnamed,
+        batch.candidates
+      )
 
       if (batch.counts.length > 0) {
         await trx.rawQuery(
-          `INSERT INTO rule_counts (range_id, rule, key, articles, files)
+          `INSERT INTO rule_counts (block_id, rule, key, articles, files)
            SELECT ?, v.rule, v.key, v.articles, v.files FROM jsonb_to_recordset(?::jsonb) AS v(rule text, key text, articles bigint, files bigint)
-           ON CONFLICT (range_id, rule, key) DO UPDATE SET articles = rule_counts.articles + excluded.articles, files = rule_counts.files + excluded.files`,
-          [rangeId, JSON.stringify(batch.counts)]
+           ON CONFLICT (block_id, rule, key) DO UPDATE SET articles = rule_counts.articles + excluded.articles, files = rule_counts.files + excluded.files`,
+          [blockId, JSON.stringify(batch.counts)]
         )
       }
 
       for (const s of batch.samples) {
         await trx.insertQuery().table('rule_samples').insert({
-          range_id: rangeId,
+          block_id: blockId,
           rule: s.rule,
           rule_key: s.ruleKey,
           release_key: s.key,
@@ -86,18 +123,6 @@ class BatchStore {
         })
       }
 
-      for (let i = 0; i < batch.fragments.length; i += 2000) {
-        const rows = batch.fragments.slice(i, i + 2000).map((f) => ({
-          range_id: rangeId,
-          file_key: f.fileKey,
-          // Not Math.min(...): a big file's fragment has more segments than the call stack takes arguments.
-          article_number: f.segments.reduce((min, s) => (s[3] < min ? s[3] : min), Infinity),
-          segments: f.segments.length,
-          fragment: JSON.stringify(f),
-        }))
-        await trx.insertQuery().table('fragments').multiInsert(rows)
-      }
-
       const moved = await trx
         .from('ranges')
         .where('id', rangeId)
@@ -107,106 +132,171 @@ class BatchStore {
         .update({
           next_id: batch.next,
           headers: db.raw('headers + ?', [batch.headers]),
-          // The range's posting time span (release_final): the batches' trimmed spans, min and max (least and greatest skip nulls).
-          posted_low: db.raw('least(posted_low, to_timestamp(?::double precision / 1000))', [batch.postedLow ?? null]),
-          posted_high: db.raw('greatest(posted_high, to_timestamp(?::double precision / 1000))', [batch.postedHigh ?? null]),
+          // The range's posting time span: the batches' trimmed spans, min and max (least and greatest skip nulls).
+          posted_low: db.raw('least(posted_low, to_timestamp(?::double precision / 1000))', [
+            batch.postedLow ?? null,
+          ]),
+          posted_high: db.raw('greatest(posted_high, to_timestamp(?::double precision / 1000))', [
+            batch.postedHigh ?? null,
+          ]),
           heartbeat_at: new Date(),
         })
-      if (Number(moved) !== 1 && !(Array.isArray(moved) && moved.length === 1)) throw new StaleBatch(`Range ${rangeId} moved on meanwhile`)
+      if (Number(moved) !== 1 && !(Array.isArray(moved) && moved.length === 1))
+        throw new StaleBatch(`Range ${rangeId} moved on meanwhile`)
       return result
     })
   }
 
-  /**
-   * An upload spread over groups (FTD: par2 and part1 to a.b.music.misc,
-   * part2 to a.b.sounds.music, each article in one group only; LOG.md
-   * 2026-10-11 (9)): a chunk without a release in its own group joins one of
-   * its key in another group whose window is within UPLOAD_GAP_MS, when none
-   * of its file names are in that release yet. A post of the same files to
-   * another group (other message-ids, the same names) stays a release of its
-   * own, or the NZB would hold every file twice. Reads the candidate's file
-   * names from its chunks; candidates are few.
-   */
-  private async joinAcrossGroups(
-    trx: TransactionClientContract,
-    part: ReleaseChunk[],
-    windows: { idx: number; key: string; first: number | null | undefined; last: number | null | undefined }[],
-    groupId: number,
-    joined: Map<number, number>
-  ) {
-    const open = windows.filter((w) => !joined.has(w.idx) && w.first)
-    if (open.length === 0) return
-    const { rows: candidates } = await trx.rawQuery(
-      `SELECT v.idx, r.id
-       FROM jsonb_to_recordset(?::jsonb) AS v(idx int, key text, first bigint, last bigint)
-       JOIN releases r ON r.identity_key = v.key AND r.group_id <> ?
-       WHERE r.posted_at <= to_timestamp((v.last + ?) / 1000.0) AND coalesce(r.last_posted_at, r.posted_at) >= to_timestamp((v.first - ?) / 1000.0)
-       ORDER BY v.idx, r.last_posted_at DESC NULLS LAST, r.id DESC`,
-      [JSON.stringify(open), groupId, UPLOAD_GAP_MS, UPLOAD_GAP_MS]
-    )
-    if (candidates.length === 0) return
-    const ids: number[] = [...new Set<number>(candidates.map((c: { id: string }) => Number(c.id)))]
-    const names = new Map<number, Set<string>>()
-    const { rows: chunks } = await trx.rawQuery(`SELECT release_id, payload FROM release_chunks WHERE release_id = ANY(?)`, [ids])
-    for (const c of chunks) {
-      const set = names.get(Number(c.release_id)) ?? new Set<string>()
-      for (const f of decodeChunk(c.payload)) set.add(f.name)
-      names.set(Number(c.release_id), set)
-    }
-    for (const c of candidates as { idx: number; id: string }[]) {
-      if (joined.has(c.idx)) continue
-      const have = names.get(Number(c.id)) ?? new Set<string>()
-      const incoming = decodeChunk(part[c.idx].payload).map((f) => f.name)
-      if (incoming.every((name) => !have.has(name))) {
-        joined.set(c.idx, Number(c.id))
-        for (const name of incoming) have.add(name)
-        names.set(Number(c.id), have)
-      }
-    }
+  /** The range's block; a range without one (no planner yet) is a block of its own, leased as the range is. */
+  async blockOf(trx: TransactionClientContract, range: any): Promise<number> {
+    if (range.block_id) return Number(range.block_id)
+    const [block] = await trx
+      .table('blocks')
+      .insert({
+        group_id: range.group_id,
+        start_id: range.start_id,
+        end_id: range.end_id,
+        next_id: range.next_id,
+        status: 'leased',
+        priority: range.priority ?? 0,
+        node_id: range.node_id,
+        leased_at: range.leased_at ?? new Date(),
+        heartbeat_at: new Date(),
+      })
+      .returning('id')
+    await trx.from('ranges').where('id', range.id).update({ block_id: block.id })
+    return Number(block.id)
   }
 
   /**
-   * Release aggregates in the catalog, one write-once chunk per release, and
-   * probe candidates (a batch's, or files the fragment merger completed).
-   * A chunk joins the release of its key whose articles the range's group
-   * owns and whose posting window is within UPLOAD_GAP_MS of its own, else
-   * starts one: re-uploads are releases of their own (decision d). The
-   * group is locked, so concurrent writers don't both start one.
+   * The unnamed releases of a batch into pending_releases (migration 6300, the HDD): keys, counts, the upload
+   * tool and the probe targets, no chunk. The pending ladder (REBUILD.md decision f) probes them cheapest first,
+   * and a name moves one into the catalog; a provisional tail release sent again adds its counts up.
+   */
+  async writePending(
+    trx: TransactionClientContract,
+    groupId: number,
+    releases: ReleaseChunk[],
+    candidates: ProbeCandidate[]
+  ): Promise<number> {
+    if (releases.length === 0) return 0
+    const targets = new Map<
+      string,
+      { kind: string; messageId: string; file: string; bytes?: number }[]
+    >()
+    for (const c of candidates) {
+      const key = slot(c.releaseKey, c.upload ?? 0)
+      const list = targets.get(key) ?? []
+      list.push({ kind: c.kind, messageId: c.messageId, file: c.file, bytes: c.bytes })
+      targets.set(key, list)
+    }
+    // In the record's lock order (LOG.md (136)): all rows sorted, then all classes sorted, in one transaction. The
+    // upsert locks a finished row too, before its WHERE passes it by; a chunk's classes before the next chunk's rows
+    // crossed the ladder's record(), which holds its rows and then takes its classes.
+    const rows = lockOrder(
+      releases.map((r) => {
+        const probes = targets.get(slot(r.key, r.upload)) ?? []
+        return {
+          release_key: releaseKey(r.key, groupId, r.firstArticle).toString('hex'),
+          identity_key: identityKey(r.key).toString('hex'),
+          name: r.name.slice(0, TEXT_MAX),
+          poster: r.poster || null,
+          first: r.postedAt,
+          last: r.postedLast ?? r.postedAt,
+          first_article: r.firstArticle,
+          last_article: r.lastArticle,
+          files: r.files,
+          segments: r.segments,
+          bytes: r.bytes,
+          partial: r.partial === true,
+          finished: r.finished !== false,
+          file_total: r.fileTotal,
+          tool: r.tool ?? (probes[0] ? uploadTool(probes[0].messageId) : 'unknown'),
+          candidates: probes,
+          // The first target's article (by effort): the cheapest probe first within a class (LOG.md (130)).
+          target_bytes: byEffort(probes)[0]?.bytes ?? null,
+        }
+      }),
+      (r) => r.release_key
+    )
+    let written = 0
+    const classes = new Map<string, { tool: string; klass: number; waiting: number }>()
+    for (let i = 0; i < rows.length; i += 1000) {
+      const { rows: done } = await trx.rawQuery(
+        `INSERT INTO pending_releases (release_key, identity_key, group_id, name, poster, posted_at, last_posted_at, first_article, last_article,
+                                       files, segments, bytes, partial, finished, file_total, tool, candidates, klass, target_bytes)
+         SELECT decode(v.release_key, 'hex'), decode(v.identity_key, 'hex'), ?, v.name, v.poster, to_timestamp(v.first / 1000.0), to_timestamp(v.last / 1000.0),
+                v.first_article, v.last_article, v.files, v.segments, v.bytes, v.partial, v.finished, v.file_total, v.tool, v.candidates,
+                pending_class(v.name, v.bytes, v.partial), v.target_bytes
+         FROM jsonb_to_recordset(?::jsonb) AS v(release_key text, identity_key text, name text, poster text, first bigint, last bigint, first_article bigint, last_article bigint,
+                                               files int, segments bigint, bytes bigint, partial boolean, finished boolean, file_total int, tool text, candidates jsonb, target_bytes bigint)
+         ON CONFLICT (release_key) DO UPDATE SET
+           files = pending_releases.files + excluded.files, segments = pending_releases.segments + excluded.segments, bytes = pending_releases.bytes + excluded.bytes,
+           last_posted_at = greatest(pending_releases.last_posted_at, excluded.last_posted_at), last_article = greatest(pending_releases.last_article, excluded.last_article),
+           partial = excluded.partial, finished = excluded.finished, file_total = coalesce(excluded.file_total, pending_releases.file_total),
+           candidates = CASE WHEN jsonb_array_length(pending_releases.candidates) = 0 THEN excluded.candidates ELSE pending_releases.candidates END,
+           klass = pending_class(pending_releases.name, pending_releases.bytes + excluded.bytes, excluded.partial),
+           target_bytes = coalesce(pending_releases.target_bytes, excluded.target_bytes)
+         WHERE NOT pending_releases.finished
+         RETURNING release_key, tool, klass, (xmax = 0) AS inserted`,
+        [groupId, JSON.stringify(rows.slice(i, i + 1000))]
+      )
+      written += done.length
+      // Each class seen gets its row in the ladder's record, so the claim explores it (pending_classes).
+      for (const d of done) {
+        const key = `${d.tool}\t${d.klass}`
+        const entry = classes.get(key) ?? { tool: d.tool, klass: Number(d.klass), waiting: 0 }
+        if (d.inserted) entry.waiting++
+        classes.set(key, entry)
+      }
+    }
+    if (classes.size > 0) {
+      await trx.rawQuery(
+        `INSERT INTO pending_classes (tool, group_id, klass, waiting)
+         SELECT v.tool, ?, v.klass, v.waiting FROM jsonb_to_recordset(?::jsonb) AS v(tool text, klass int, waiting int)
+         ON CONFLICT (tool, group_id, klass) DO UPDATE SET waiting = pending_classes.waiting + excluded.waiting`,
+        [
+          groupId,
+          JSON.stringify(lockOrder(classes.values(), (c) => `${c.tool}\t${groupId}\t${c.klass}`)),
+        ]
+      )
+    }
+    return written
+  }
+
+  /**
+   * The releases of a batch into the catalog, one chunk each, and their probe candidates. Every chunk is one
+   * release: inserted, or (a provisional tail release sent again) finished.
    */
   async writeReleases(
     trx: TransactionClientContract,
-    rangeId: number,
+    where: { groupId: number; blockId: number },
     batchFrom: number,
     releases: ReleaseChunk[],
     candidates: ProbeCandidate[]
   ) {
-    const out = { newReleases: 0, candidates: 0 }
+    const out = { newReleases: 0, finalized: 0, candidates: 0 }
+    if (releases.length === 0) return out
+    /** Release ids by slot, for the chunks and candidates; a slot not here was a finished release sent again. */
     const ids = new Map<string, number>()
-    const slot = (key: string, upload: number) => `${key}\n${upload}`
-    const { group_id: groupId } = await trx.from('ranges').where('id', rangeId).select('group_id').firstOrFail()
-    // One writer per group at a time: a chunk only matches releases of its group, and a lock per key
-    // ran out of shared memory (a picture group's batch has 50k keys; LOG.md 2026-10-09 (12)).
-    await trx.rawQuery(`SELECT pg_advisory_xact_lock(hashtextextended('releases:' || ?::text, 0))`, [groupId])
+    const finalized = new Set<number>()
+
+    const posters = await CatalogRefs.posterIds(trx, releases.map((r) => r.poster).filter(Boolean))
+    const groupSets = await CatalogRefs.groupSetIds(
+      trx,
+      releases.map((r) => r.groups)
+    )
     const sorted = [...releases].sort((a, b) => a.key.localeCompare(b.key) || a.upload - b.upload)
     for (let i = 0; i < sorted.length; i += 1000) {
       const part = sorted.slice(i, i + 1000)
-      const windows = part.map((r, idx) => ({ idx, key: r.key, first: r.postedAt, last: r.postedLast ?? r.postedAt }))
-      const { rows: matches } = await trx.rawQuery(
-        `SELECT DISTINCT ON (v.idx) v.idx, r.id
-         FROM jsonb_to_recordset(?::jsonb) AS v(idx int, key text, first bigint, last bigint)
-         JOIN releases r ON r.identity_key = v.key AND r.group_id = ?
-         WHERE v.first IS NULL OR r.posted_at IS NULL
-            OR (r.posted_at <= to_timestamp((v.last + ?) / 1000.0) AND coalesce(r.last_posted_at, r.posted_at) >= to_timestamp((v.first - ?) / 1000.0))
-         ORDER BY v.idx, r.last_posted_at DESC NULLS LAST, r.id DESC`,
-        [JSON.stringify(windows), groupId, UPLOAD_GAP_MS, UPLOAD_GAP_MS]
-      )
-      const joined = new Map<number, number>(matches.map((m: { idx: number; id: string }) => [m.idx, Number(m.id)]))
-      await this.joinAcrossGroups(trx, part, windows, groupId, joined)
-      const row = (r: ReleaseChunk) => ({
+      const rows = part.map((r, idx) => ({
+        idx,
+        release_key: releaseKey(r.key, where.groupId, r.firstArticle).toString('hex'),
+        identity_key: identityKey(r.key).toString('hex'),
         name: r.name,
-        source: r.source,
-        poster: r.poster,
-        groups: r.groups,
+        source: sourceCode(r.source),
+        poster_id: r.poster ? (posters.get(posterText(r.poster)) ?? null) : null,
+        group_set_id: groupSets.get(groupSetKey(r.groups))!,
         first: r.postedAt,
         last: r.postedLast ?? r.postedAt,
         file_total: r.fileTotal,
@@ -217,107 +307,114 @@ class BatchStore {
         last_article: r.lastArticle,
         readable: r.readable,
         has_par2: r.hasPar2,
-      })
-      const updates = part.flatMap((r, idx) => (joined.has(idx) ? [{ id: joined.get(idx)!, ...row(r) }] : []))
-      if (updates.length > 0) {
-        await trx.rawQuery(
-          `WITH v AS (SELECT * FROM jsonb_to_recordset(?::jsonb) AS v(id bigint, first bigint, last bigint, file_total int, files int, segments bigint, bytes bigint, first_article bigint, last_article bigint, readable boolean, has_par2 boolean)),
-                a AS (SELECT id, sum(files) AS files, sum(segments) AS segments, sum(bytes) AS bytes, count(*) AS chunks, min(first) AS first, max(last) AS last,
-                             max(file_total) AS file_total, min(first_article) AS first_article, max(last_article) AS last_article, bool_or(readable) AS readable, bool_or(has_par2) AS has_par2
-                      FROM v GROUP BY id)
-           UPDATE releases r SET
-             files = r.files + a.files,
-             segments = r.segments + a.segments,
-             bytes = r.bytes + a.bytes,
-             chunks = r.chunks + a.chunks,
-             posted_at = least(r.posted_at, to_timestamp(a.first / 1000.0)),
-             last_posted_at = greatest(r.last_posted_at, to_timestamp(a.last / 1000.0)),
-             file_total = coalesce(r.file_total, a.file_total),
-             first_article = least(r.first_article, a.first_article),
-             last_article = greatest(r.last_article, a.last_article),
-             readable = r.readable OR a.readable,
-             has_par2 = r.has_par2 OR a.has_par2,
-             -- As it stands (the counter's files all found); settling makes it final. Without a counter, what its par2
-             -- set decided stays (app/services/par2_completeness.ts).
-             complete = CASE WHEN coalesce(r.file_total, a.file_total) IS NULL THEN r.complete ELSE r.files + a.files >= coalesce(r.file_total, a.file_total) END,
-             last_seen_at = now()
-           FROM a WHERE r.id = a.id`,
-          [JSON.stringify(updates)]
+        finished: r.finished !== false,
+        edge: r.edge === true,
+        partial: r.partial === true,
+        sidecar_only: sidecarOnly(r),
+      }))
+      // Inserted, or a provisional release of the key finished (the chunk's files added up); a finished release
+      // sent again (no row returned) is left as it is.
+      const { rows: written } = await trx.rawQuery(
+        `INSERT INTO releases (release_key, identity_key, group_id, group_set_id, poster_id, name, source, posted_at, last_posted_at, file_total, files, segments, bytes, chunks,
+                               first_article, last_article, readable, has_par2, complete, edge, sidecar_only, settled_at)
+         SELECT decode(v.release_key, 'hex'), decode(v.identity_key, 'hex'), ?, v.group_set_id, v.poster_id, v.name, v.source,
+                to_timestamp(v.first / 1000.0), to_timestamp(v.last / 1000.0), v.file_total, v.files, v.segments, v.bytes, 1,
+                v.first_article, v.last_article, v.readable, v.has_par2,
+                CASE WHEN v.partial THEN false WHEN v.file_total IS NULL THEN NULL ELSE v.files >= v.file_total END,
+                v.edge, v.sidecar_only, CASE WHEN v.finished THEN now() END
+         FROM jsonb_to_recordset(?::jsonb) AS v(idx int, release_key text, identity_key text, name text, source int, poster_id int, group_set_id int, first bigint, last bigint,
+                                               file_total int, files int, segments bigint, bytes bigint, first_article bigint, last_article bigint, readable boolean, has_par2 boolean,
+                                               finished boolean, edge boolean, partial boolean, sidecar_only boolean)
+         ON CONFLICT (release_key) DO UPDATE SET
+           files = releases.files + excluded.files, segments = releases.segments + excluded.segments, bytes = releases.bytes + excluded.bytes,
+           chunks = releases.chunks + CASE WHEN excluded.files > 0 THEN 1 ELSE 0 END,
+           posted_at = least(releases.posted_at, excluded.posted_at), last_posted_at = greatest(releases.last_posted_at, excluded.last_posted_at),
+           file_total = coalesce(excluded.file_total, releases.file_total),
+           first_article = least(releases.first_article, excluded.first_article), last_article = greatest(releases.last_article, excluded.last_article),
+           readable = releases.readable OR excluded.readable, has_par2 = releases.has_par2 OR excluded.has_par2,
+           complete = CASE WHEN releases.complete IS FALSE AND releases.file_total IS NULL THEN false
+                           WHEN coalesce(excluded.file_total, releases.file_total) IS NULL THEN releases.complete
+                           ELSE releases.files + excluded.files >= coalesce(excluded.file_total, releases.file_total) END,
+           edge = excluded.edge, sidecar_only = coalesce(releases.sidecar_only, excluded.sidecar_only), settled_at = excluded.settled_at
+         WHERE releases.settled_at IS NULL
+         RETURNING id, encode(release_key, 'hex') AS release_key, (xmax = 0) AS inserted, settled_at`,
+        [where.groupId, JSON.stringify(rows)]
+      )
+      const byKey = new Map<string, { id: number; inserted: boolean; settled: boolean }>(
+        written.map(
+          (w: { id: string; release_key: string; inserted: boolean; settled_at: Date | null }) => [
+            w.release_key,
+            { id: Number(w.id), inserted: w.inserted, settled: w.settled_at !== null },
+          ]
         )
-        part.forEach((r, idx) => joined.has(idx) && ids.set(slot(r.key, r.upload), joined.get(idx)!))
-      }
-      const fresh = part.filter((_, idx) => !joined.has(idx))
-      if (fresh.length > 0) {
-        // The first upload of a key keeps the key; later ones are key:2, key:3, … after the highest number: merges
-        // delete releases, and a count fell behind it (83 releases of a key up to key:129: 20 numbers tried, all
-        // taken, the range's batch failed for good; LOG.md 2026-10-20 (89)). Spot releases (spot:key) aren't numbered.
-        const { rows: counts } = await trx.rawQuery(
-          `SELECT identity_key, max(CASE WHEN release_key = identity_key THEN 1
-                                         WHEN left(release_key, length(identity_key) + 1) = identity_key || ':' AND substr(release_key, length(identity_key) + 2) ~ '^[0-9]{1,9}$'
-                                         THEN substr(release_key, length(identity_key) + 2)::int END) AS n
-           FROM releases WHERE identity_key = ANY(?) GROUP BY 1`,
-          [[...new Set(fresh.map((r) => r.key))]]
-        )
-        const taken = new Map<string, number>(counts.map((c: { identity_key: string; n: string }) => [c.identity_key, Number(c.n)]))
-        const numbered = (r: ReleaseChunk) => {
-          const n = (taken.get(r.key) ?? 0) + 1
-          taken.set(r.key, n)
-          return { release_key: n === 1 ? r.key : `${r.key}:${n}`, key: r.key, upload: r.upload, ...row(r), sidecar_only: sidecarOnly(r) }
+      )
+      const texts: { release_id: number; kind: number; text: string }[] = []
+      for (const row of rows) {
+        const w = byKey.get(row.release_key)
+        if (!w) continue
+        const r = part[row.idx]
+        ids.set(slot(r.key, r.upload), w.id)
+        if (w.inserted) out.newReleases++
+        else if (w.settled) {
+          out.finalized++
+          finalized.add(w.id)
         }
-        // Their posters, for poster search (posters: one row per poster, searched by trigram; migration 4000).
-        const posters = [...new Set(fresh.map((r) => r.poster).filter((p): p is string => !!p))]
-        if (posters.length > 0) await trx.rawQuery(`INSERT INTO posters (poster) SELECT DISTINCT left(p, 200) FROM unnest(?::text[]) AS p ON CONFLICT DO NOTHING`, [posters])
-        let inserts = fresh.map(numbered)
-        // Writers of other groups number the same key at the same time (a spam post crossposted to 67
-        // groups; only the group is locked): a number taken meanwhile is skipped, and the next one tried.
-        for (let attempt = 0; inserts.length > 0; attempt++) {
-          if (attempt === 20) throw new Error(`No free release key for ${inserts[0].key}`)
-          const { rows } = await trx.rawQuery(
-            `INSERT INTO releases (release_key, identity_key, group_id, name, source, poster, groups, posted_at, last_posted_at, file_total, files, segments, bytes, chunks, first_article, last_article, readable, has_par2, complete, sidecar_only)
-             SELECT v.release_key, v.key, ?, v.name, v.source, v.poster, v.groups, to_timestamp(v.first / 1000.0), to_timestamp(v.last / 1000.0), v.file_total, v.files, v.segments, v.bytes, 1,
-                    v.first_article, v.last_article, v.readable, v.has_par2, CASE WHEN v.file_total IS NULL THEN NULL ELSE v.files >= v.file_total END, v.sidecar_only
-             FROM jsonb_to_recordset(?::jsonb) AS v(release_key text, key text, name text, source text, poster text, groups jsonb, first bigint, last bigint, file_total int, files int, segments bigint, bytes bigint, first_article bigint, last_article bigint, readable boolean, has_par2 boolean, sidecar_only boolean)
-             ON CONFLICT (release_key) DO NOTHING
-             RETURNING id, release_key`,
-            [groupId, JSON.stringify(inserts)]
-          )
-          const byReleaseKey = new Map<string, number>(rows.map((x: { id: string; release_key: string }) => [x.release_key, Number(x.id)]))
-          const missed: typeof inserts = []
-          for (const r of inserts) {
-            const id = byReleaseKey.get(r.release_key)
-            if (id === undefined) missed.push(r)
-            else ids.set(slot(r.key, r.upload), id)
+        texts.push({ release_id: w.id, kind: TEXT_KINDS.name, text: row.name.slice(0, TEXT_MAX) })
+        // An album's members stay findable by their own names (the first TEXTS_PER_KIND of them).
+        if (r.source === 'album' && r.files > 0) {
+          for (const text of textsToStore(decodeChunk(r.payload).map((f) => f.name))) {
+            texts.push({ release_id: w.id, kind: TEXT_KINDS.member, text })
           }
-          out.newReleases += rows.length
-          const chunks = new Map(fresh.map((r) => [slot(r.key, r.upload), r]))
-          inserts = missed.map((r) => numbered(chunks.get(slot(r.key, r.upload))!))
         }
       }
+      // The name (and an album's members), searchable (release_texts; a text sent again is a no-op).
+      if (texts.length > 0) {
+        await trx.rawQuery(
+          `INSERT INTO release_texts (release_id, kind, text)
+           SELECT v.release_id, v.kind, v.text FROM jsonb_to_recordset(?::jsonb) AS v(release_id bigint, kind int, text text) WHERE v.text <> ''
+           ON CONFLICT DO NOTHING`,
+          [JSON.stringify(texts)]
+        )
+      }
     }
 
-    // One write-once chunk per release.
+    // One write-once chunk per release (a finish adds the files not sent before).
     for (let i = 0; i < sorted.length; i += 500) {
-      const rows = sorted.slice(i, i + 500).map((r) => ({
-        release_id: ids.get(slot(r.key, r.upload))!,
-        range_id: rangeId,
-        batch_from: batchFrom,
-        files: r.files,
-        segments: r.segments,
-        payload: r.payload,
-      }))
+      const rows = sorted
+        .slice(i, i + 500)
+        .filter((r) => ids.has(slot(r.key, r.upload)) && r.files > 0)
+        .map((r) => ({
+          release_id: ids.get(slot(r.key, r.upload))!,
+          block_id: where.blockId,
+          batch_from: batchFrom,
+          files: r.files,
+          segments: r.segments,
+          payload: r.payload,
+        }))
       if (rows.length > 0) await trx.insertQuery().table('release_chunks').multiInsert(rows)
     }
 
+    // A finished release's candidates are claimable at once; a provisional one's wait for its finish.
+    const settled = new Set<number>()
+    for (const r of sorted) {
+      const id = ids.get(slot(r.key, r.upload))
+      if (id !== undefined && r.finished !== false) settled.add(id)
+    }
+    const chunks = new Map(sorted.map((r) => [slot(r.key, r.upload), r]))
     const kept = candidates.filter((c) => ids.has(slot(c.releaseKey, c.upload ?? 0)))
     for (let i = 0; i < kept.length; i += 2000) {
-      const rows = kept
-        .slice(i, i + 2000)
-        .map((c) => ({
-          release_id: ids.get(slot(c.releaseKey, c.upload ?? 0))!,
+      const rows = kept.slice(i, i + 2000).map((c) => {
+        const releaseId = ids.get(slot(c.releaseKey, c.upload ?? 0))!
+        const chunk = chunks.get(slot(c.releaseKey, c.upload ?? 0))!
+        return {
+          release_id: releaseId,
           kind: c.kind,
           message_id: c.messageId,
           file: c.file,
-        }))
+          ready: settled.has(releaseId),
+          unreadable: !chunk.readable,
+        }
+      })
       const added = await trx
         .knexQuery()
         .table('probe_candidates')
@@ -329,15 +426,14 @@ class BatchStore {
       // A later piece of a split 7z: the tails of lower volumes go (app/services/seven_zip_tails.ts).
       const tails = new Set(rows.filter((r) => r.kind === '7z-tail').map((r) => r.release_id))
       if (added.length > 0 && tails.size > 0) await pruneSevenZipTails(trx, [...tails])
-      // Files added to a settled release (a later range of its upload): the settler made its candidates claimable
-      // already and won't again (93,642 waited since 2026-10-13; LOG.md 2026-10-20 (87)).
-      if (added.length > 0) {
-        await trx.rawQuery(
-          `UPDATE probe_candidates pc SET ready = true, unreadable = NOT r.readable FROM releases r
-           WHERE pc.id = ANY(?) AND r.id = pc.release_id AND r.settled_at IS NOT NULL`,
-          [added.map((a: { id: string | number }) => String(a.id))]
-        )
-      }
+    }
+    // The finish of a provisional release: its waiting candidates become claimable.
+    if (finalized.size > 0) {
+      await trx.rawQuery(
+        `UPDATE probe_candidates pc SET ready = true, unreadable = NOT r.readable FROM releases r
+         WHERE pc.release_id = ANY(?) AND r.id = pc.release_id AND pc.status = 'waiting' AND NOT pc.ready`,
+        [[...finalized]]
+      )
     }
 
     return out

app/services/catalog_refs.ts

@@ -0,0 +1,93 @@
+import type db from '@adonisjs/lucid/services/db'
+
+type Client = Pick<typeof db, 'rawQuery'>
+
+/** Posters are kept to this length (as the trigram index and poster search always did; a btree entry holds 2,704 bytes). */
+export const POSTER_MAX = 200
+/** Cached ids per process: group sets are ~2,600 in all; posters many more, so the cache is emptied when it grows past this. */
+const CACHE_MAX = 200_000
+
+const posterCache = new Map<string, number>()
+const groupSetCache = new Map<string, number>()
+
+/** A group list as group_sets stores it: sorted, without repeats, as JSON. */
+export function groupSetKey(groups: string[]): string {
+  return JSON.stringify([...new Set(groups)].sort())
+}
+
+/** A poster as posters stores it. */
+export function posterText(poster: string): string {
+  return poster.slice(0, POSTER_MAX)
+}
+
+async function ids(
+  client: Client,
+  cache: Map<string, number>,
+  table: 'posters' | 'group_sets',
+  column: 'poster' | 'groups',
+  cast: 'text' | 'jsonb',
+  values: string[]
+): Promise<Map<string, number>> {
+  const out = new Map<string, number>()
+  const missing: string[] = []
+  for (const v of values) {
+    const id = cache.get(v)
+    if (id === undefined) missing.push(v)
+    else out.set(v, id)
+  }
+  if (missing.length === 0) return out
+  // Insert what's new, then read every id (ON CONFLICT DO NOTHING returns nothing for the rows that existed).
+  await client.rawQuery(
+    `INSERT INTO ${table} (${column}) SELECT DISTINCT v::${cast} FROM unnest(?::text[]) AS v ON CONFLICT DO NOTHING`,
+    [missing]
+  )
+  const { rows } = await client.rawQuery(
+    `SELECT id, ${column}::text AS value FROM ${table} WHERE ${column} = ANY(?::${cast}[])`,
+    [missing]
+  )
+  if (cache.size + rows.length > CACHE_MAX) cache.clear()
+  for (const r of rows) {
+    // jsonb comes back in Postgres' spelling ("["a", "b"]"); the caller's key is the JSON we sent.
+    const key = cast === 'jsonb' ? JSON.stringify(JSON.parse(r.value)) : r.value
+    cache.set(key, Number(r.id))
+    out.set(key, Number(r.id))
+  }
+  return out
+}
+
+/**
+ * Ids in the catalog's lookup tables (voyager 3, migration 6000): a release's poster_id and group_set_id. New
+ * values are inserted; known ones come from a cache per process (the batch store sees the same posters and
+ * group sets batch after batch).
+ */
+export default {
+  /** Ids by poster (posterText applied), for a batch of posters. */
+  async posterIds(client: Client, posters: string[]): Promise<Map<string, number>> {
+    return ids(client, posterCache, 'posters', 'poster', 'text', [
+      ...new Set(posters.map(posterText)),
+    ])
+  },
+
+  async posterId(client: Client, poster: string): Promise<number> {
+    const found = await this.posterIds(client, [poster])
+    return found.get(posterText(poster))!
+  },
+
+  /** Ids by groupSetKey, for a batch of group lists. */
+  async groupSetIds(client: Client, sets: string[][]): Promise<Map<string, number>> {
+    return ids(client, groupSetCache, 'group_sets', 'groups', 'jsonb', [
+      ...new Set(sets.map(groupSetKey)),
+    ])
+  },
+
+  async groupSetId(client: Client, groups: string[]): Promise<number> {
+    const found = await this.groupSetIds(client, [groups])
+    return found.get(groupSetKey(groups))!
+  },
+
+  /** Forgets the cached ids (tests roll their transactions back, so the ids they saw are gone). */
+  forget() {
+    posterCache.clear()
+    groupSetCache.clear()
+  },
+}

app/services/corpus.ts

@@ -42,13 +42,13 @@ class Corpus {
     return coordinatorConfig.corpus.dir
   }
 
-  private rangeDir(range: CorpusRange) {
-    return join(this.dir!, range.group, `${range.start}-${range.end}`)
+  private rangeDir(range: CorpusRange, dir = this.dir!) {
+    return join(dir, range.group, `${range.start}-${range.end}`)
   }
 
-  private path(range: CorpusRange, span: ChunkSpan) {
+  private path(range: CorpusRange, span: ChunkSpan, dir = this.dir!) {
     const index = Math.floor((span.from - range.start) / coordinatorConfig.chunkArticles)
-    return join(this.rangeDir(range), `${String(index).padStart(4, '0')}-${span.from}-${span.to}.gz`)
+    return join(this.rangeDir(range, dir), `${String(index).padStart(4, '0')}-${span.from}-${span.to}.gz`)
   }
 
   /** Whether every chunk of the range is stored (its range.json). */
@@ -69,10 +69,10 @@ class Corpus {
     )
   }
 
-  /** A stored chunk, or null. */
-  async read(range: CorpusRange, span: ChunkSpan): Promise<Buffer | null> {
-    if (!this.dir) return null
-    return readFile(this.path(range, span)).catch(() => null)
+  /** A stored chunk, or null (`dir`: another corpus than CORPUS_DIR's, the namer's tests). */
+  async read(range: CorpusRange, span: ChunkSpan, dir = this.dir): Promise<Buffer | null> {
+    if (!dir) return null
+    return readFile(this.path(range, span, dir)).catch(() => null)
   }
 
   /**

app/services/fragment_merger.ts

@@ -1,153 +0,0 @@
-import db from '@adonisjs/lucid/services/db'
-import BatchStore from '#services/batch_store'
-import ChunkPartitions from '#services/chunk_partitions'
-import RuleSets from '#services/rule_sets'
-import { RuleSet } from '#lib/rules/index'
-import { candidatesOf, chunkOf, mergeFragments, releaseKeyOf, uploadsOf, type Fragment, type ProbeCandidate, type ReleaseChunk } from '#lib/edge/index'
-import type { AggregatedFile } from '#lib/files/index'
-import type { EdgeSegment } from '#lib/edge/index'
-
-export interface MergeResult {
-  files: number
-  completed: number
-  incomplete: number
-  waiting: number
-  /** This round reached the end of the file keys (the next one starts over). */
-  wrapped: boolean
-}
-
-/** File keys per round. */
-const BATCH = 20_000
-
-/**
- * Files cut by a range's edge (REBUILD.md, fragments), put together once
- * the ranges around them are done: a complete file is judged by the rules
- * like any other and written as a release chunk (with probe candidates); a
- * file still missing parts while every neighbouring range is done is
- * incomplete, and dropped (counted on its ranges). Others wait.
- */
-class FragmentMerger {
-  /** Where the next round goes on: waiting files must not take every round's places. */
-  private cursor = ''
-
-  async run(): Promise<MergeResult> {
-    const result: MergeResult = { files: 0, completed: 0, incomplete: 0, waiting: 0, wrapped: false }
-    const { rows: keys } = await db.rawQuery(
-      `SELECT DISTINCT f.file_key FROM fragments f JOIN ranges r ON r.id = f.range_id
-       WHERE r.status = 'done' AND f.file_key > ? ORDER BY f.file_key LIMIT ?`,
-      [this.cursor, BATCH]
-    )
-    // At the end of the keys: the next round starts over.
-    result.wrapped = keys.length < BATCH
-    this.cursor = result.wrapped ? '' : keys.at(-1).file_key
-    if (keys.length === 0) return result
-    const rules = new RuleSet(await RuleSets.current())
-    const rows = await db
-      .from('fragments as f')
-      .join('ranges as r', 'r.id', 'f.range_id')
-      .whereIn('f.file_key', keys.map((k: { file_key: string }) => k.file_key))
-      .select('f.id', 'f.file_key', 'f.range_id', 'f.fragment', 'r.group_id', 'r.start_id', 'r.end_id', 'r.status')
-    const byKey = new Map<string, typeof rows>()
-    for (const row of rows) {
-      const list = byKey.get(row.file_key)
-      if (list) list.push(row)
-      else byKey.set(row.file_key, [row])
-    }
-
-    const completed: { file: AggregatedFile<EdgeSegment>; rangeId: number; ids: number[] }[] = []
-    const incomplete: { rangeId: number; ids: number[]; segments: number }[] = []
-    const neighbours = await this.neighbourStatus([...byKey.values()])
-    for (const list of byKey.values()) {
-      result.files++
-      const file = mergeFragments(list.map((r) => r.fragment as Fragment))
-      const ids = list.map((r) => Number(r.id))
-      const rangeId = Math.max(...list.map((r) => Number(r.range_id)))
-      if (file) {
-        completed.push({ file, rangeId, ids })
-        continue
-      }
-      if (list.every((r) => r.status === 'done') && neighbours(list)) {
-        incomplete.push({ rangeId, ids, segments: list.reduce((n, r) => n + (r.fragment as Fragment).segments.length, 0) })
-      } else {
-        result.waiting++
-      }
-    }
-
-    // Completed files, grouped into their releases, judged and written like a batch's.
-    const byRelease = new Map<string, { files: AggregatedFile<EdgeSegment>[]; rangeId: number }>()
-    for (const { file, rangeId } of completed) {
-      const key = releaseKeyOf(file.segments[0])
-      const entry = byRelease.get(key)
-      if (entry) entry.files.push(file)
-      else byRelease.set(key, { files: [file], rangeId })
-    }
-    const releases: { rangeId: number; chunk: ReleaseChunk; candidates: ProbeCandidate[] }[] = []
-    for (const [key, { files, rangeId }] of byRelease) {
-      const verdict = rules.judge(key, files.map((f) => ({ name: f.segments[0].file.name, totalParts: f.segments[0].totalParts, groups: f.segments[0].file.groups, firstMessageId: f.segments[0].messageId, poster: f.segments[0].file.poster })))
-      if (verdict.rule !== null) continue
-      uploadsOf(files).forEach((list, upload) => releases.push({ rangeId, chunk: chunkOf(key, list, upload), candidates: candidatesOf(key, list, upload) }))
-    }
-    await ChunkPartitions.ensure()
-    await db.transaction(async (trx) => {
-      // In key order (batches lock their group; the merger only ever waits on one batch at a time).
-      releases.sort((a, b) => a.chunk.key.localeCompare(b.chunk.key) || a.chunk.upload - b.chunk.upload)
-      for (const { rangeId, chunk, candidates } of releases) await BatchStore.writeReleases(trx, rangeId, 0, [chunk], candidates)
-      const done = [...completed.flatMap((c) => c.ids), ...incomplete.flatMap((c) => c.ids)]
-      for (let i = 0; i < done.length; i += 5000) await trx.from('fragments').whereIn('id', done.slice(i, i + 5000)).delete()
-      for (const { rangeId, segments } of incomplete) {
-        await trx.rawQuery(
-          `INSERT INTO rule_counts (range_id, rule, key, articles, files) VALUES (?, 'incomplete', 'edge', ?, 1)
-           ON CONFLICT (range_id, rule, key) DO UPDATE SET articles = rule_counts.articles + excluded.articles, files = rule_counts.files + 1`,
-          [rangeId, segments]
-        )
-      }
-    })
-    result.completed = completed.length
-    result.incomplete = incomplete.length
-    return result
-  }
-
-  /**
-   * For each file's fragments: whether the ranges right before and after
-   * their ranges (where a missing part could be) are done, read in one
-   * query for the whole round (per file it was three, ~2,000 files a
-   * minute). A neighbour not planned yet counts as not done (newer articles
-   * come later), except before the group's oldest article.
-   */
-  private async neighbourStatus(lists: { group_id: number; start_id: number | string; end_id: number | string }[][]) {
-    const span = (list: (typeof lists)[number]) => ({
-      group: Number(list[0].group_id),
-      first: Math.min(...list.map((r) => Number(r.start_id))),
-      last: Math.max(...list.map((r) => Number(r.end_id))),
-    })
-    const spans = lists.map(span)
-    const groups = [...new Set(spans.map((s) => s.group))]
-    // The range holding the article before and the one after, each by the (group_id, start_id) index:
-    // "end_id = ? OR start_id = ?" read every range of the group per file (5.5 s a round; LOG.md 2026-10-09).
-    const { rows } = await db.rawQuery(
-      `SELECT s.group_id, s.before_end, s.after_start, b.status AS before, a.status AS after
-       FROM unnest(?::int[], ?::bigint[], ?::bigint[]) AS s(group_id, before_end, after_start)
-       LEFT JOIN LATERAL (SELECT r.status FROM ranges r WHERE r.group_id = s.group_id AND r.start_id <= s.before_end AND r.end_id >= s.before_end
-                          ORDER BY r.start_id DESC LIMIT 1) b ON true
-       LEFT JOIN LATERAL (SELECT r.status FROM ranges r WHERE r.group_id = s.group_id AND r.start_id <= s.after_start AND r.end_id >= s.after_start
-                          ORDER BY r.start_id DESC LIMIT 1) a ON true`,
-      [spans.map((s) => s.group), spans.map((s) => s.first - 1), spans.map((s) => s.last + 1)]
-    )
-    const byEnd = new Map<string, string>()
-    const byStart = new Map<string, string>()
-    for (const r of rows) {
-      if (r.before !== null) byEnd.set(`${r.group_id}:${r.before_end}`, r.before)
-      if (r.after !== null) byStart.set(`${r.group_id}:${r.after_start}`, r.after)
-    }
-    const lows = new Map((await db.from('groups').whereIn('id', groups).select('id', 'low')).map((g) => [Number(g.id), g.low === null ? null : Number(g.low)]))
-    return (list: (typeof lists)[number]) => {
-      const { group, first, last } = span(list)
-      const before = byEnd.get(`${group}:${first - 1}`)
-      const low = lows.get(group)
-      const beforeDone = before !== undefined ? before === 'done' : low === null || low === undefined || first <= low
-      return beforeDone && byStart.get(`${group}:${last + 1}`) === 'done'
-    }
-  }
-}
-
-export default new FragmentMerger()

app/services/group_leases.ts

@@ -0,0 +1,280 @@
+import db from '@adonisjs/lucid/services/db'
+import { createReadStream, createWriteStream } from 'node:fs'
+import { mkdir, rename, stat } from 'node:fs/promises'
+import { join } from 'node:path'
+import type { Readable } from 'node:stream'
+import { pipeline } from 'node:stream/promises'
+import coordinatorConfig from '#config/coordinator'
+import RuleSets from '#services/rule_sets'
+import CatalogSpace from '#services/catalog_space'
+import { LeaseLost } from '#services/leases'
+import type { GroupLease, Lane, NextRange } from '#lib/edge/index'
+
+export const LANES: Lane[] = ['history', 'tail']
+
+/**
+ * Group leases (REBUILD.md, "Voyager 3: window indexing", 1 and 5): a worker thread holds a group and asks for its
+ * next range each time it finished one, so a group's ranges are indexed in article order through one carried
+ * window. The coordinator keeps one row a group and lane (group_leases): the holder, its heartbeat, the range it
+ * works and the snapshot its window stands at.
+ *
+ * - Two lanes (2026-10-27): the `history` lane works the ranges the bounds hand out (LEASE_MIN/MAX_PRIORITY,
+ *   LEASE_CAPTURE_ONLY: the windows, the history), the `tail` lane the tail planner's ranges (priority at or below
+ *   the tail's). Each lane has its own holder, window and snapshot, so a thread on the tail never wades through a
+ *   group's history first (a group's ranges come in article order, and a held group's history is days of ranges), and
+ *   the tail of a group whose history is held isn't stuck behind it. The one seam a group is where the history lane
+ *   reaches the tail lane's first range (its ranges are done: the history lane is caught up).
+ * - lease: the free group with the most urgent planned range of the lane (the ranges' priority), and among those a
+ *   mix: the group with the most planned articles and the longest unheld one (never held first) take turns, so a
+ *   big group snapshotted at a stop doesn't wait behind hundreds of small never-held ones, and a quiet group's tail
+ *   is visited all the same. A holder quiet for LEASE_TTL_SECONDS is taken over. Taking a group resets what the
+ *   lane's snapshot doesn't cover: the lane's ranges from the snapshot's article on that the window did after it
+ *   (leased, done since the snapshot, or read past it) are planned again (the batch store's keys make the redo
+ *   idempotent); without a snapshot the window is lost and the group continues at its resume point (one seam).
+ * - next: the group's lowest planned range of the lane (or the one this node already has), leased to the node; none
+ *   when the lane is caught up: the worker snapshots and gives the group back, and gets it again when the tail
+ *   planned more.
+ * - snapshot: the serialized window, kept as a file on bulk (SNAPSHOT_DIR; `<group>.gz`, the tail's
+ *   `<group>-tail.gz`), given to the next holder of the lane.
+ */
+class GroupLeases {
+  get dir(): string {
+    return coordinatorConfig.snapshotDir
+  }
+
+  /**
+   * The lane's ranges, as a condition on `r`: the tail lane's are at or below the tail's priority, the history lane's
+   * above it — within the lease bounds when `bounds` (what is handed out), or all of them (what a lane's window did).
+   */
+  private laneFilter(lane: Lane, bounds: boolean): { sql: string; bindings: number[] } {
+    const tail = coordinatorConfig.tail.priority
+    if (lane === 'tail') return { sql: 'r.priority <= ?', bindings: [tail] }
+    if (!bounds) return { sql: 'r.priority > ?', bindings: [tail] }
+    const { leasePriorityMin: min, leasePriorityMax: max, leaseCaptureOnly } = coordinatorConfig
+    return {
+      sql: `r.priority > ? AND r.priority BETWEEN ? AND ?${leaseCaptureOnly ? ' AND r.capture' : ''}`,
+      bindings: [tail, min, max],
+    }
+  }
+
+  async lease(nodeId: number, lane: Lane = 'history'): Promise<GroupLease | null> {
+    // The catalog at its cap on the NVMe: the tail only (CatalogSpace), as the range leases did.
+    if (lane === 'history' && (await CatalogSpace.full())) return null
+    const rules = await RuleSets.current()
+    const ttl = coordinatorConfig.leaseTtlSeconds
+    const todo = this.laneFilter(lane, true)
+    const all = this.laneFilter(lane, false)
+    return db.transaction(async (trx) => {
+      // One lease at a time: the choice reads several tables.
+      await trx.rawQuery(`SELECT pg_advisory_xact_lock(hashtextextended('group-lease', 0))`)
+      // The free groups with planned work in the lane (a range leased by a holder gone quiet counts: a group whose
+      // last range that holder took would never be taken over), by the most urgent range's priority; within a priority the
+      // most planned articles and the longest unheld alternate: one lease starts from the biggest, the next from
+      // the longest unheld (the sequence counts the leases; ranked afresh each time, "biggest first" would always
+      // win).
+      const { rows: turns } = await trx.rawQuery(`SELECT nextval('group_lease_turns') % 2 AS odd`)
+      const [first, second] = Number(turns[0].odd) === 1 ? [1, 0] : [0, 1]
+      const { rows } = await trx.rawQuery(
+        `WITH todo AS (
+           SELECT r.group_id, min(r.priority) AS priority, sum(r.end_id - r.next_id + 1) AS articles
+           FROM ranges r
+           WHERE (r.status = 'planned' OR (r.status = 'leased' AND r.heartbeat_at < now() - make_interval(secs => ?)))
+             AND r.next_id <= r.end_id AND ${todo.sql}
+           GROUP BY r.group_id
+         ), free AS (
+           SELECT g.id, g.name, g.priority AS group_priority, l.snapshot_next, l.snapshot_at, l.heartbeat_at,
+                  todo.priority, todo.articles
+           FROM todo JOIN groups g ON g.id = todo.group_id AND g.active
+           LEFT JOIN group_leases l ON l.group_id = g.id AND l.lane = ?
+           WHERE l.node_id IS NULL OR l.heartbeat_at < now() - make_interval(secs => ?)
+         ), turns AS (
+           SELECT free.*,
+                  least(2 * row_number() OVER (PARTITION BY priority ORDER BY articles DESC, group_priority, id) - ?,
+                        2 * row_number() OVER (PARTITION BY priority ORDER BY heartbeat_at NULLS FIRST, group_priority, id) - ?) AS turn
+           FROM free
+         )
+         SELECT * FROM turns ORDER BY priority, turn, group_priority, id LIMIT 1`,
+        [ttl, ...todo.bindings, lane, ttl, first, second]
+      )
+      const row = rows[0]
+      if (!row) return null
+      const groupId = Number(row.id)
+      // What the snapshot doesn't cover is read again: the lane's ranges from its article on that the window did
+      // after it (a range done before the snapshot was taken — by the old range leases, say — stays done).
+      if (row.snapshot_next !== null) {
+        await trx.rawQuery(
+          `UPDATE ranges r SET status = 'planned', node_id = NULL, next_id = greatest(r.start_id, ?::bigint)
+           WHERE r.group_id = ? AND r.end_id >= ?::bigint AND ${all.sql}
+             AND (r.status = 'leased' OR (r.status = 'done' AND (r.done_at IS NULL OR r.done_at >= ?)) OR (r.status = 'planned' AND r.next_id > ?::bigint))`,
+          [
+            row.snapshot_next,
+            groupId,
+            row.snapshot_next,
+            ...all.bindings,
+            row.snapshot_at ?? new Date(0),
+            row.snapshot_next,
+          ]
+        )
+      } else {
+        await trx.rawQuery(
+          `UPDATE ranges r SET status = 'planned', node_id = NULL WHERE r.group_id = ? AND r.status = 'leased' AND ${all.sql}`,
+          [groupId, ...all.bindings]
+        )
+      }
+      await trx.rawQuery(
+        `INSERT INTO group_leases (group_id, lane, node_id, leased_at, heartbeat_at, range_id)
+         VALUES (?, ?, ?, now(), now(), NULL)
+         ON CONFLICT (group_id, lane) DO UPDATE SET node_id = excluded.node_id, leased_at = now(), heartbeat_at = now(), range_id = NULL`,
+        [groupId, lane, nodeId]
+      )
+      await trx.from('nodes').where('id', nodeId).update({ last_seen_at: new Date() })
+      return {
+        group: { id: groupId, name: row.name, priority: Number(row.group_priority) },
+        lane,
+        rules,
+        snapshot: row.snapshot_next !== null ? { next: Number(row.snapshot_next) } : null,
+      }
+    })
+  }
+
+  /** The group's lease row of the lane, when this node holds it. */
+  async held(groupId: number, nodeId: number, lane: Lane = 'history') {
+    const lease = await db
+      .from('group_leases')
+      .where('group_id', groupId)
+      .where('lane', lane)
+      .first()
+    if (!lease || Number(lease.node_id) !== nodeId)
+      throw new LeaseLost(`Group ${groupId} (${lane}) is not held by this node`)
+    return lease
+  }
+
+  /**
+   * The group's next range of the lane in article order (or the one this node has), leased to the node; null when
+   * the lane is caught up.
+   */
+  async next(groupId: number, nodeId: number, lane: Lane = 'history'): Promise<NextRange | null> {
+    await this.held(groupId, nodeId, lane)
+    const rules = await RuleSets.current()
+    const todo = this.laneFilter(lane, true)
+    const all = this.laneFilter(lane, false)
+    const mine = (
+      await db.rawQuery(
+        `SELECT r.id, r.start_id, r.end_id, r.next_id FROM ranges r
+         WHERE r.group_id = ? AND r.status = 'leased' AND r.node_id = ? AND ${all.sql} ORDER BY r.start_id LIMIT 1`,
+        [groupId, nodeId, ...all.bindings]
+      )
+    ).rows[0]
+    const row =
+      mine ??
+      (
+        await db.rawQuery(
+          `UPDATE ranges r SET status = 'leased', node_id = ?, leased_at = now(), heartbeat_at = now(), attempts = attempts + 1, rule_version = ?
+           WHERE r.id = (
+             SELECT r.id FROM ranges r WHERE r.group_id = ? AND r.status = 'planned' AND r.next_id <= r.end_id AND ${todo.sql}
+             ORDER BY r.start_id LIMIT 1 FOR UPDATE SKIP LOCKED)
+           RETURNING r.id, r.start_id, r.end_id, r.next_id`,
+          [nodeId, rules.version, groupId, ...todo.bindings]
+        )
+      ).rows[0]
+    await db
+      .from('group_leases')
+      .where('group_id', groupId)
+      .where('lane', lane)
+      .update({ range_id: row ? Number(row.id) : null, heartbeat_at: new Date() })
+    if (!row) return null
+    return {
+      range: {
+        id: Number(row.id),
+        start: Number(row.start_id),
+        end: Number(row.end_id),
+        next: Number(row.next_id),
+      },
+    }
+  }
+
+  async heartbeat(groupId: number, nodeId: number, lane: Lane = 'history') {
+    await this.held(groupId, nodeId, lane)
+    await db
+      .from('group_leases')
+      .where('group_id', groupId)
+      .where('lane', lane)
+      .update({ heartbeat_at: new Date() })
+    await db.from('nodes').where('id', nodeId).update({ last_seen_at: new Date() })
+  }
+
+  /** The lane a range belongs to, by its priority. */
+  laneOf(priority: number): Lane {
+    return priority <= coordinatorConfig.tail.priority ? 'tail' : 'history'
+  }
+
+  /** A range of the group finished by its holder: counted on the lane's lease (Leases.finish does the range). */
+  async rangeDone(groupId: number, headers: number, lane: Lane = 'history') {
+    await db.rawQuery(
+      `UPDATE group_leases SET ranges_done = ranges_done + 1, headers = headers + ?, range_id = NULL WHERE group_id = ? AND lane = ?`,
+      [headers, groupId, lane]
+    )
+  }
+
+  /** The worker gives the group back (after its snapshot): its range stays at its resume point for the next holder. */
+  async release(groupId: number, nodeId: number, lane: Lane = 'history') {
+    await this.held(groupId, nodeId, lane)
+    const all = this.laneFilter(lane, false)
+    await db.rawQuery(
+      `UPDATE ranges r SET status = 'planned', node_id = NULL
+       WHERE r.group_id = ? AND r.status = 'leased' AND r.node_id = ? AND ${all.sql}`,
+      [groupId, nodeId, ...all.bindings]
+    )
+    await db
+      .from('group_leases')
+      .where('group_id', groupId)
+      .where('lane', lane)
+      .update({ node_id: null, range_id: null })
+  }
+
+  snapshotPath(groupId: number, lane: Lane = 'history') {
+    return join(this.dir, lane === 'tail' ? `${groupId}-tail.gz` : `${groupId}.gz`)
+  }
+
+  /**
+   * The window's snapshot standing at `next` (the next article to index), streamed to disk (a busy group's is
+   * gigabytes) and written whole before it replaces the old one.
+   */
+  async putSnapshot(
+    groupId: number,
+    nodeId: number,
+    next: number,
+    body: Readable,
+    lane: Lane = 'history'
+  ) {
+    await this.held(groupId, nodeId, lane)
+    await mkdir(this.dir, { recursive: true })
+    const path = this.snapshotPath(groupId, lane)
+    await pipeline(body, createWriteStream(`${path}.part`))
+    await rename(`${path}.part`, path)
+    await db
+      .from('group_leases')
+      .where('group_id', groupId)
+      .where('lane', lane)
+      .update({ snapshot_next: next, snapshot_at: new Date(), heartbeat_at: new Date() })
+  }
+
+  /** The snapshot's stream and size, or null when there is none. */
+  async getSnapshot(
+    groupId: number,
+    nodeId: number,
+    lane: Lane = 'history'
+  ): Promise<{ stream: Readable; size: number; next: number } | null> {
+    const lease = await this.held(groupId, nodeId, lane)
+    if (lease.snapshot_next === null) return null
+    const path = this.snapshotPath(groupId, lane)
+    try {
+      const { size } = await stat(path)
+      return { stream: createReadStream(path), size, next: Number(lease.snapshot_next) }
+    } catch {
+      return null
+    }
+  }
+}
+
+export default new GroupLeases()

app/services/hash_twins.ts

@@ -0,0 +1,36 @@
+import db from '@adonisjs/lucid/services/db'
+
+/**
+ * Twins by content (LOG.md (130)): a probed file's first 16 KiB and size (file_hashes, what the par2 merge and the
+ * repeated verdict match) in a catalog release that has a name, readable or found inside. The same film posted
+ * under scrambles by other tools gets that name without another fetch. Releases with a junk verdict don't count.
+ */
+/** Twins read per pair: the hottest pairs have 131k. */
+const TWIN_SCAN = 500
+
+export async function twinNames(
+  pairs: { hash16k: string; size: number }[],
+  trx: { rawQuery: typeof db.rawQuery } = db
+): Promise<Map<string, string>> {
+  const out = new Map<string, string>()
+  if (pairs.length === 0) return out
+  // Per pair the first named twin among its lowest TWIN_SCAN release ids. hash16k is typed as the column
+  // (char(32)): as text the join cast the column and scanned all of file_hashes, 4 s a call; and with the index
+  // the hottest pairs (55 with over 500 twins, up to 131k, all readable) made a call walk every twin, 37 s inside
+  // the flush's transaction (LOG.md (138)). One name is all a pair needs.
+  const { rows } = await trx.rawQuery(
+    `SELECT v.hash16k, v.size, t.name
+     FROM jsonb_to_recordset(?::jsonb) AS v(hash16k char(32), size bigint)
+     CROSS JOIN LATERAL (
+       SELECT coalesce(r.inner_name, r.name) AS name
+       FROM (SELECT release_id FROM file_hashes h WHERE h.hash16k = v.hash16k AND h.size = v.size ORDER BY release_id LIMIT ?) h
+       JOIN releases r ON r.id = h.release_id AND (r.readable OR r.inner_name IS NOT NULL)
+       WHERE NOT EXISTS (SELECT 1 FROM release_verdicts x WHERE x.release_id = r.id)
+       ORDER BY r.readable DESC, r.id LIMIT 1) t`,
+    [JSON.stringify(pairs), TWIN_SCAN]
+  )
+  for (const r of rows) out.set(`${r.hash16k}\t${r.size}`, r.name)
+  return out
+}
+
+export const twinKey = (hash16k: string, size: number) => `${hash16k}\t${size}`

app/services/nzb_service.ts

@@ -2,6 +2,7 @@ import { gunzipSync } from 'node:zlib'
 import db from '@adonisjs/lucid/services/db'
 import { decodeChunk, type ChunkFile } from '#lib/chunks/index'
 import { buildNzb } from '#lib/nzb/index'
+import { SOURCES } from '#lib/release/source'
 
 /** Where a known password came from, the most trusted first: the spot's NZB meta, its title, its text. */
 const PASSWORD_SOURCES = ['nzb', 'spot-title', 'spot-text']
@@ -35,7 +36,7 @@ class NzbService {
     if (!release) return null
     const password = (await this.passwords(releaseId))[0]?.password
     // A release a spot made of its own (no files matched the header pass's): the spot's NZB as read.
-    if (release.source === 'spot') {
+    if (Number(release.source) === SOURCES.spot) {
       const spot = await db.from('spots').where('release_id', releaseId).whereNotNull('nzb').select('nzb').first()
       if (spot) return { name: release.name, xml: withPassword(gunzipSync(spot.nzb).toString('utf8'), password) }
     }

app/services/par2_merger.ts

@@ -93,11 +93,12 @@ const HOLDERS = `SELECT c.release_id FROM (
 
 interface Member {
   id: number
-  release_key: string
   group_id: number
   files: number
   posted_at: Date | null
   readable: boolean
+  poster_id?: number | null
+  /** The poster's text (joined from posters) where a query needs it. */
   poster?: string | null
   file_total?: number | null
 }
@@ -295,7 +296,8 @@ export class Par2Merger {
    */
   async nameMatches(trx: any, setId: string): Promise<Map<string, Member> | null> {
     const { rows: all } = await trx.rawQuery(
-      `SELECT r.id, r.release_key, r.group_id, r.files, r.posted_at, r.readable, r.poster FROM par2_sets s JOIN releases r ON r.id = s.release_id
+      `SELECT r.id, r.group_id, r.files, r.posted_at, r.readable, r.poster_id, p.poster
+       FROM par2_sets s JOIN releases r ON r.id = s.release_id LEFT JOIN posters p ON p.id = r.poster_id
        WHERE s.set_id = ? ORDER BY r.posted_at DESC NULLS LAST, r.id`,
       [setId]
     )
@@ -310,12 +312,12 @@ export class Par2Merger {
     const names = new Set(described.map((d: { name: string }) => fileKey(d.name)))
     const at = new Date(anchor.posted_at).getTime()
     const { rows: candidates } = await trx.rawQuery(
-      `SELECT r.id, r.release_key, r.group_id, r.files, r.posted_at, r.readable, r.poster FROM releases r
-       WHERE left(r.poster, 200) = left(?, 200) AND r.poster = ? AND r.group_id = ? AND r.posted_at BETWEEN ? AND ? AND r.id <> ALL(?)
+      `SELECT r.id, r.group_id, r.files, r.posted_at, r.readable, r.poster_id FROM releases r
+       WHERE r.poster_id = ? AND r.group_id = ? AND r.posted_at BETWEEN ? AND ? AND r.id <> ALL(?)
          AND r.files <= ?
-       ORDER BY left(r.poster, 200), r.posted_at DESC NULLS LAST, r.id DESC LIMIT ?`,
+       ORDER BY r.posted_at DESC NULLS LAST, r.id DESC LIMIT ?`,
       // More files than the set describes, with its sidecars, can't all be named in it.
-      [anchor.poster, anchor.poster, anchor.group_id, new Date(at - NAME_SPREAD_MS), new Date(at + NAME_SPREAD_MS), holders.map((h: Member) => h.id), described.length + SLACK_FILES, NAME_CANDIDATES]
+      [anchor.poster_id, anchor.group_id, new Date(at - NAME_SPREAD_MS), new Date(at + NAME_SPREAD_MS), holders.map((h: Member) => h.id), described.length + SLACK_FILES, NAME_CANDIDATES]
     )
     const found = new Map<string, Member>(holders.map((h: Member) => [String(h.id), h]))
     let files = holders.reduce((n: number, h: Member) => n + Number(h.files), 0)
@@ -344,7 +346,7 @@ export class Par2Merger {
   async mergeStem(stem: string): Promise<number> {
     return db.transaction(async (trx) => {
       const { rows } = await trx.rawQuery(
-        `SELECT DISTINCT r.id, r.release_key, r.group_id, r.files, r.posted_at, r.readable, r.poster, r.file_total
+        `SELECT DISTINCT r.id, r.group_id, r.files, r.posted_at, r.readable, r.poster_id, r.file_total
          FROM release_inspections i JOIN releases r ON r.id = i.release_id
          WHERE i.kind = 'lone' AND ${INNER_STEM} = ?`,
         [stem]
@@ -356,7 +358,7 @@ export class Par2Merger {
       const total = (r: Member) => (r.file_total ? Number(r.file_total) : null)
       const near = (r: Member) =>
         r.group_id === anchor.group_id &&
-        (r.poster ?? '') === (anchor.poster ?? '') &&
+        (r.poster_id ?? null) === (anchor.poster_id ?? null) &&
         (total(r) === null || total(anchor) === null || total(r) === total(anchor)) &&
         (at(r) === null || at(anchor) === null || Math.abs(at(r)! - at(anchor)!) <= MAX_SPREAD_MS)
       const found = new Map<string, Member>()
@@ -374,7 +376,7 @@ export class Par2Merger {
   async mergeSet(setId: string): Promise<number> {
     return db.transaction(async (trx) => {
       const { rows: holders } = await trx.rawQuery(
-        `SELECT r.id, r.release_key, r.group_id, r.files, r.posted_at, r.readable FROM par2_sets s JOIN releases r ON r.id = s.release_id
+        `SELECT r.id, r.group_id, r.files, r.posted_at, r.readable FROM par2_sets s JOIN releases r ON r.id = s.release_id
          WHERE s.set_id = ? ORDER BY r.id`,
         [setId]
       )
@@ -383,7 +385,7 @@ export class Par2Merger {
       const at = (r: Member) => (r.posted_at ? new Date(r.posted_at).getTime() : null)
       const near = (r: Member) => r.group_id === anchor.group_id && (at(r) === null || at(anchor) === null || Math.abs(at(r)! - at(anchor)!) <= MAX_SPREAD_MS)
       const { rows: matches } = await trx.rawQuery(
-        `SELECT f.file_id, r.id, r.release_key, r.group_id, r.files, r.posted_at, r.readable
+        `SELECT f.file_id, r.id, r.group_id, r.files, r.posted_at, r.readable
          FROM par2_files f CROSS JOIN LATERAL (${HOLDERS}) h JOIN releases r ON r.id = h.release_id
          WHERE f.set_id = ?`,
         [setId]
@@ -426,7 +428,7 @@ export class Par2Merger {
 
       await trx.rawQuery(
         `INSERT INTO release_merges (from_id, from_key, into_id, set_id, files, via)
-         SELECT id, release_key, ?, ?, files, ? FROM releases WHERE id = ANY(?)`,
+         SELECT id, encode(release_key, 'hex'), ?, ?, files, ? FROM releases WHERE id = ANY(?)`,
         [target.id, key, via, others]
       )
       // A candidate of an article the target, or another member of lower id, has already (the same file in two releases)
@@ -462,13 +464,13 @@ export class Par2Merger {
            files = t.files + s.files, segments = t.segments + s.segments, bytes = t.bytes + s.bytes, chunks = t.chunks + s.chunks,
            first_article = least(t.first_article, s.first_article), last_article = greatest(t.last_article, s.last_article),
            posted_at = least(t.posted_at, s.posted_at), last_posted_at = greatest(t.last_posted_at, s.last_posted_at),
-           first_seen_at = least(t.first_seen_at, s.first_seen_at), last_seen_at = greatest(t.last_seen_at, s.last_seen_at),
+           first_seen_at = least(t.first_seen_at, s.first_seen_at),
            readable = t.readable OR s.readable, has_par2 = t.has_par2 OR s.has_par2, passworded = t.passworded OR s.passworded,
            file_total = greatest(t.file_total, s.file_total), inner_name = coalesce(t.inner_name, s.inner_name),
            name = coalesce(nullif(?, ''), t.name)
          FROM (SELECT sum(files) AS files, sum(segments) AS segments, sum(bytes) AS bytes, sum(chunks) AS chunks,
                  min(first_article) AS first_article, max(last_article) AS last_article, min(posted_at) AS posted_at,
-                 max(last_posted_at) AS last_posted_at, min(first_seen_at) AS first_seen_at, max(last_seen_at) AS last_seen_at,
+                 max(last_posted_at) AS last_posted_at, min(first_seen_at) AS first_seen_at,
                  bool_or(readable) AS readable, bool_or(has_par2) AS has_par2, bool_or(passworded) AS passworded,
                  max(file_total) AS file_total, max(inner_name) AS inner_name
                FROM releases WHERE id = ANY(?)) s

app/services/par2_pruner.ts

@@ -75,7 +75,7 @@ export class Par2Pruner {
                   WHERE o.set_id = b.set_id AND v.release_id IS NULL) THEN 'judged'
            WHEN NOT EXISTS (SELECT 1 FROM par2_sets s WHERE s.set_id = b.set_id AND s.created_at > ?)
             AND NOT EXISTS (SELECT 1 FROM par2_sets s JOIN releases r ON r.id = s.release_id WHERE s.set_id = b.set_id AND (
-                  release_final(r.group_id, r.first_article, r.last_article, r.posted_at, r.last_posted_at, r.complete) IS NOT TRUE
+                  r.settled_at IS NULL
                   OR EXISTS (SELECT 1 FROM recent x WHERE x.group_id = r.group_id
                     AND x.posted_low <= coalesce(r.last_posted_at, r.posted_at) + interval '6 hours' AND x.posted_high >= r.posted_at - interval '6 hours')))
             AND NOT EXISTS (SELECT 1 FROM par2_files f JOIN unseen u ON u.hash16k = f.hash16k AND u.size = f.length WHERE f.set_id = b.set_id)

app/services/pending_ladder.ts

@@ -0,0 +1,514 @@
+import db from '@adonisjs/lucid/services/db'
+import type { TransactionClientContract } from '@adonisjs/lucid/types/database'
+import coordinatorConfig from '#config/coordinator'
+import { explored } from '#lib/rules/index'
+import { usableName, type Inspection } from '#lib/probe/index'
+import RuleSets from '#services/rule_sets'
+import { twinKey, twinNames } from '#services/hash_twins'
+
+/** pending_releases.status (migration 6400). */
+export const PENDING = {
+  waiting: 0,
+  named: 1,
+  discarded: 2,
+  exhausted: 3,
+  naming: 4,
+  taken: 9,
+} as const
+
+/** The probe targets of a pending release in the order of effort: a par2 names the set's files outright, an archive's start its entries, a lone file only its yEnc name, a 7z's tail is the costliest find. */
+export const EFFORT = ['par2', 'archive', 'lone', '7z-tail'] as const
+
+export interface PendingTarget {
+  kind: string
+  messageId: string
+  file: string
+  /** The target article's bytes (candidates from 2026-11-02 on). */
+  bytes?: number
+}
+
+/** One probe the ladder hands the prober: the next target of a pending release. */
+export interface PendingProbe {
+  /** pending_releases.release_key as hex. */
+  releaseKey: string
+  kind: string
+  messageId: string
+  file: string
+  tool: string
+  groupId: number
+  /** Its class (lib/release/classes.ts; 0 for a row from before the classes). */
+  klass: number
+  poster: string
+  /** Targets probed before this one. */
+  tries: number
+  targets: number
+}
+
+export type PendingOutcome = {
+  probe: PendingProbe
+  status: 'done' | 'missing' | 'error' | 'waiting'
+  inspection?: Inspection
+}
+
+export function byEffort(targets: PendingTarget[]): PendingTarget[] {
+  const rank = (kind: string) => {
+    const i = (EFFORT as readonly string[]).indexOf(kind)
+    return i < 0 ? EFFORT.length : i
+  }
+  return [...targets].sort((a, b) => rank(a.kind) - rank(b.kind))
+}
+
+/** A class's record (pending_classes), for the claim's order. */
+interface ClassRow {
+  tool: string
+  groupId: number
+  klass: number
+  probed: number
+  named: number
+  waiting: number
+}
+
+/** The record is read again this often. */
+const CLASSES_MS = 60_000
+/** A class found short of a claim's share is left alone this long. */
+const EMPTY_MS = 5 * 60_000
+/**
+ * Rows in one order for a statement that locks many of them (pending_releases by key, pending_classes and
+ * pending_signatures by theirs): the batch store, the record, the claim and the sweep all lock the same way, so two
+ * of them can't deadlock on each other (LOG.md (134)).
+ */
+export function lockOrder<T>(rows: Iterable<T>, key: (row: T) => string): T[] {
+  return [...rows].sort((a, b) => {
+    const ka = key(a)
+    const kb = key(b)
+    return ka < kb ? -1 : ka > kb ? 1 : 0
+  })
+}
+
+/** The prior rate of an unexplored class and its weight in probes: the ladder's overall rate on 2026-11-02 (LOG.md (130)). */
+const PRIOR_RATE = 0.15
+const PRIOR_WEIGHT = 10
+/** One class takes at most this share of a claim: the rest explores the classes after it. */
+const CLASS_SHARE = 0.25
+
+/** The expected naming rate of a class: its record shrunk towards the prior. */
+export function classScore(c: { probed: number; named: number }): number {
+  return (c.named + PRIOR_RATE * PRIOR_WEIGHT) / (c.probed + PRIOR_WEIGHT)
+}
+
+/** An UPDATE returns its rows in no order: the claim's (the cheapest target, then the oldest) restored. */
+function claimed(rows: any[]): any[] {
+  const bytes = (r: any) =>
+    r.target_bytes === null ? Number.POSITIVE_INFINITY : Number(r.target_bytes)
+  return [...rows].sort(
+    (a, b) =>
+      bytes(a) - bytes(b) ||
+      new Date(a.created_at).getTime() - new Date(b.created_at).getTime() ||
+      Number(a.first_article) - Number(b.first_article)
+  )
+}
+
+function toProbe(r: any): PendingProbe {
+  const targets = byEffort(r.candidates as PendingTarget[])
+  const tries = Number(r.tries)
+  const target = targets[Math.min(tries, targets.length - 1)]
+  return {
+    releaseKey: r.release_key,
+    kind: target.kind,
+    messageId: target.messageId,
+    file: target.file,
+    tool: r.tool,
+    groupId: Number(r.group_id),
+    klass: Number(r.klass ?? 0),
+    poster: r.poster,
+    tries,
+    targets: targets.length,
+  }
+}
+
+/**
+ * The pending ladder (REBUILD.md decision f): pending_releases probed from least to most effort, a usable name
+ * handing the release to the namer (app/services/pending_namer.ts), no name after the last target exhausting it;
+ * every finished climb counted on its signature (tool, group, poster; and the tool and group together) and on its
+ * class (tool, group, size bucket × name shape × partial; lib/release/classes.ts, LOG.md (130)), and a signature
+ * or class with enough probes and next to no names dropped: its waiting releases are discarded without a probe,
+ * a sample of them (the rules' samplePercent) still climbing so the rate keeps being measured. The claim takes the
+ * classes with the best record first (classScore), the cheapest target first within one.
+ */
+class PendingLadder {
+  /** A previous prober's claims back in the queue. */
+  async release(): Promise<number> {
+    const { rowCount } = await db.rawQuery(
+      `UPDATE pending_releases SET status = ? WHERE status = ?`,
+      [PENDING.waiting, PENDING.taken]
+    )
+    return rowCount ?? 0
+  }
+
+  /** The ladder's record per class, cached CLASSES_MS between claims, best first; classes found empty, until when. */
+  private classes: ClassRow[] = []
+  private classesAt = 0
+  private empty = new Map<string, number>()
+
+  /** The record read again on the next claim (the tests). */
+  forget() {
+    this.classesAt = 0
+    this.empty.clear()
+  }
+
+  /** pending_classes with waiting releases and not dropped, by their expected rate (a prior for the unexplored). */
+  private async classRows(): Promise<ClassRow[]> {
+    if (Date.now() - this.classesAt < CLASSES_MS) return this.classes
+    const { rows } = await db.rawQuery(
+      `SELECT tool, group_id, klass, probed, named, waiting FROM pending_classes WHERE waiting > 0 AND NOT dropped`
+    )
+    this.classes = rows
+      .map((r: any) => ({
+        tool: r.tool as string,
+        groupId: Number(r.group_id),
+        klass: Number(r.klass),
+        probed: Number(r.probed),
+        named: Number(r.named),
+        waiting: Number(r.waiting),
+      }))
+      .sort((a: ClassRow, b: ClassRow) => classScore(b) - classScore(a))
+    this.classesAt = Date.now()
+    return this.classes
+  }
+
+  /**
+   * The next target of up to `n` waiting releases with a target left, claimed: by class, the best record first,
+   * at most CLASS_SHARE of the claim from one class (the rest explores the classes after it), the cheapest target
+   * and then the oldest within one; what the classes don't fill, the oldest of the rest (rows from before the
+   * classes, status 0 of a class without a row).
+   */
+  async claim(n: number): Promise<PendingProbe[]> {
+    if (n <= 0) return []
+    const out: PendingProbe[] = []
+    const perClass = Math.max(1, Math.floor(n * CLASS_SHARE))
+    const now = Date.now()
+    for (const c of await this.classRows()) {
+      if (out.length >= n) break
+      const key = `${c.tool}\t${c.groupId}\t${c.klass}`
+      if ((this.empty.get(key) ?? 0) > now) continue
+      const want = Math.min(perClass, n - out.length)
+      const { rows } = await db.rawQuery(
+        `UPDATE pending_releases p SET status = ? FROM (
+           SELECT release_key FROM pending_releases
+           WHERE status = ? AND tool = ? AND group_id = ? AND klass = ? AND finished AND tries < jsonb_array_length(candidates)
+           ORDER BY target_bytes NULLS LAST, created_at, first_article LIMIT ? FOR UPDATE SKIP LOCKED) c
+         WHERE p.release_key = c.release_key
+         RETURNING encode(p.release_key, 'hex') AS release_key, p.candidates, p.tries, p.tool, p.group_id, p.klass, coalesce(p.poster, '') AS poster, p.target_bytes, p.created_at, p.first_article`,
+        [PENDING.taken, PENDING.waiting, c.tool, c.groupId, c.klass, want]
+      )
+      if (rows.length < want) this.empty.set(key, now + EMPTY_MS)
+      if (rows.length > 0) {
+        await db.rawQuery(
+          `UPDATE pending_classes SET waiting = greatest(0, waiting - ?) WHERE tool = ? AND group_id = ? AND klass = ?`,
+          [rows.length, c.tool, c.groupId, c.klass]
+        )
+      }
+      out.push(...claimed(rows).map(toProbe))
+    }
+    if (out.length < n) {
+      const { rows } = await db.rawQuery(
+        `UPDATE pending_releases p SET status = ? FROM (
+           SELECT release_key FROM pending_releases
+           WHERE status = ? AND finished AND tries < jsonb_array_length(candidates)
+           ORDER BY created_at, first_article LIMIT ? FOR UPDATE SKIP LOCKED) c
+         WHERE p.release_key = c.release_key
+         RETURNING encode(p.release_key, 'hex') AS release_key, p.candidates, p.tries, p.tool, p.group_id, p.klass, coalesce(p.poster, '') AS poster, p.target_bytes, p.created_at, p.first_article`,
+        [PENDING.taken, PENDING.waiting, n - out.length]
+      )
+      out.push(...claimed(rows).map(toProbe))
+    }
+    return out
+  }
+
+  /**
+   * What the probes showed: a usable name moves the release to the namer (status 4), else the next target, or the
+   * end of the ladder (status 3); the finished climbs counted on their signatures.
+   */
+  async record(outcomes: PendingOutcome[], trx: TransactionClientContract | typeof db = db) {
+    if (outcomes.length === 0) return
+    // A twin by content names what the probe itself didn't (hash_twins.ts).
+    const twins = await twinNames(
+      outcomes
+        .filter((o) => o.status === 'done' && o.inspection?.hash16k && o.inspection.size)
+        .map((o) => ({ hash16k: o.inspection!.hash16k!, size: o.inspection!.size })),
+      trx
+    )
+    const rows = outcomes.map((o) => {
+      const p = o.probe
+      if (o.status === 'waiting') {
+        return {
+          release_key: p.releaseKey,
+          status: PENDING.waiting,
+          tries: p.tries,
+          named: null,
+          inspection: null,
+          final: null,
+        }
+      }
+      const probed = o.status === 'done' && o.inspection ? usableName(o.inspection, p.kind) : null
+      const twin =
+        o.status === 'done' && o.inspection?.hash16k
+          ? twins.get(twinKey(o.inspection.hash16k, o.inspection.size))
+          : undefined
+      const name = probed ?? twin ?? null
+      const tries = p.tries + 1
+      const inspection = {
+        kind: p.kind,
+        status: o.status,
+        result: o.inspection ? { ...o.inspection, par2: undefined } : null,
+      }
+      if (name)
+        return {
+          release_key: p.releaseKey,
+          status: PENDING.naming,
+          tries: 0,
+          named: name,
+          inspection,
+          final: 'named',
+        }
+      if (tries >= p.targets)
+        return {
+          release_key: p.releaseKey,
+          status: PENDING.exhausted,
+          tries,
+          named: null,
+          inspection,
+          final: 'exhausted',
+        }
+      return {
+        release_key: p.releaseKey,
+        status: PENDING.waiting,
+        tries,
+        named: null,
+        inspection,
+        final: null,
+      }
+    })
+    // The rows locked in key order first: the UPDATE's join below takes them in whatever order the plan does.
+    await trx.rawQuery(
+      `SELECT 1 FROM pending_releases WHERE release_key = ANY(SELECT decode(k, 'hex') FROM unnest(?::text[]) AS k)
+       ORDER BY release_key FOR UPDATE`,
+      [rows.map((r) => r.release_key).sort()]
+    )
+    await trx.rawQuery(
+      `UPDATE pending_releases p SET status = v.status, tries = v.tries, named = coalesce(v.named, p.named),
+         inspections = CASE WHEN v.inspection IS NULL THEN p.inspections ELSE p.inspections || v.inspection END,
+         probed_at = CASE WHEN v.inspection IS NULL THEN p.probed_at ELSE now() END
+       FROM jsonb_to_recordset(?::jsonb) AS v(release_key text, status int, tries int, named text, inspection jsonb)
+       WHERE p.release_key = decode(v.release_key, 'hex')`,
+      [
+        JSON.stringify(
+          lockOrder(
+            rows.map(({ final: _final, ...row }) => row),
+            (r) => r.release_key
+          )
+        ),
+      ]
+    )
+    // The signatures: the poster's own and the tool's and group's together.
+    const counts = new Map<
+      string,
+      { tool: string; group_id: number; poster: string; probed: number; named: number }
+    >()
+    for (const [i, o] of outcomes.entries()) {
+      const final = rows[i].final
+      if (!final) continue
+      for (const poster of [o.probe.poster, '']) {
+        const key = `${o.probe.tool}\t${o.probe.groupId}\t${poster}`
+        const entry = counts.get(key) ?? {
+          tool: o.probe.tool,
+          group_id: o.probe.groupId,
+          poster,
+          probed: 0,
+          named: 0,
+        }
+        entry.probed++
+        if (final === 'named') entry.named++
+        counts.set(key, entry)
+      }
+    }
+    if (counts.size === 0) return
+    const { pendingMinProbes, pendingMaxNamed } = coordinatorConfig.probe
+    await trx.rawQuery(
+      `INSERT INTO pending_signatures (tool, group_id, poster, probed, named)
+       SELECT v.tool, v.group_id, v.poster, v.probed, v.named FROM jsonb_to_recordset(?::jsonb) AS v(tool text, group_id int, poster text, probed int, named int)
+       ON CONFLICT (tool, group_id, poster) DO UPDATE SET probed = pending_signatures.probed + excluded.probed, named = pending_signatures.named + excluded.named, updated_at = now()`,
+      [JSON.stringify(lockOrder(counts.values(), (c) => `${c.tool}\t${c.group_id}\t${c.poster}`))]
+    )
+    await trx.rawQuery(
+      `UPDATE pending_signatures s SET dropped = (s.probed >= ?::int AND s.named::float <= s.probed * ?::float),
+         swept_at = CASE WHEN s.dropped THEN s.swept_at END
+       FROM jsonb_to_recordset(?::jsonb) AS v(tool text, group_id int, poster text)
+       WHERE s.tool = v.tool AND s.group_id = v.group_id AND s.poster = v.poster`,
+      [
+        pendingMinProbes,
+        pendingMaxNamed,
+        JSON.stringify(
+          lockOrder(counts.values(), (c) => `${c.tool}\t${c.group_id}\t${c.poster}`).map(
+            ({ tool, group_id, poster }) => ({ tool, group_id, poster })
+          )
+        ),
+      ]
+    )
+    // The classes: the same record per (tool, group, class), and a probe given back is waiting again.
+    const classes = new Map<
+      string,
+      {
+        tool: string
+        group_id: number
+        klass: number
+        probed: number
+        named: number
+        waiting: number
+      }
+    >()
+    for (const [i, o] of outcomes.entries()) {
+      const key = `${o.probe.tool}\t${o.probe.groupId}\t${o.probe.klass}`
+      const entry = classes.get(key) ?? {
+        tool: o.probe.tool,
+        group_id: o.probe.groupId,
+        klass: o.probe.klass,
+        probed: 0,
+        named: 0,
+        waiting: 0,
+      }
+      const final = rows[i].final
+      if (final) entry.probed++
+      if (final === 'named') entry.named++
+      if (o.status === 'waiting') entry.waiting++
+      classes.set(key, entry)
+    }
+    await trx.rawQuery(
+      `INSERT INTO pending_classes (tool, group_id, klass, probed, named, waiting)
+       SELECT v.tool, v.group_id, v.klass, v.probed, v.named, v.waiting FROM jsonb_to_recordset(?::jsonb) AS v(tool text, group_id int, klass int, probed int, named int, waiting int)
+       ON CONFLICT (tool, group_id, klass) DO UPDATE SET probed = pending_classes.probed + excluded.probed, named = pending_classes.named + excluded.named,
+         waiting = pending_classes.waiting + excluded.waiting, updated_at = now()`,
+      [JSON.stringify(lockOrder(classes.values(), (c) => `${c.tool}\t${c.group_id}\t${c.klass}`))]
+    )
+    await trx.rawQuery(
+      `UPDATE pending_classes c SET dropped = (c.probed >= ?::int AND c.named::float <= c.probed * ?::float),
+         swept_at = CASE WHEN c.dropped THEN c.swept_at END
+       FROM jsonb_to_recordset(?::jsonb) AS v(tool text, group_id int, klass int)
+       WHERE c.tool = v.tool AND c.group_id = v.group_id AND c.klass = v.klass`,
+      [
+        pendingMinProbes,
+        pendingMaxNamed,
+        JSON.stringify(
+          lockOrder(classes.values(), (c) => `${c.tool}\t${c.group_id}\t${c.klass}`).map(
+            ({ tool, group_id, klass }) => ({ tool, group_id, klass })
+          )
+        ),
+      ]
+    )
+  }
+
+  /**
+   * Waiting releases of a dropped class or signature discarded without a probe, the explored sample excepted; how
+   * many. Each dropped class or signature is swept once (swept_at; null again when the record drops it anew),
+   * `perSweep` of them at most, each by its index (the claim's for a class, the signature index for a poster's
+   * rows, the ladder's for a tool's and group's) in rounds of `limit`. The prober runs this beside the lanes, not
+   * in the claim (LOG.md (133)); the edge drops new releases of dropped signatures and classes itself, so one
+   * sweep is all a drop needs.
+   */
+  async discard(limit = 10_000, perSweep = 20): Promise<number> {
+    const { samplePercent } = await RuleSets.current()
+    let discarded = 0
+    const { rows: classes } = await db.rawQuery(
+      `SELECT tool, group_id, klass FROM pending_classes WHERE dropped AND swept_at IS NULL ORDER BY group_id, tool, klass LIMIT ?`,
+      [perSweep]
+    )
+    for (const c of classes) {
+      let rows: any[]
+      do {
+        ;({ rows } = await db.rawQuery(
+          `SELECT encode(release_key, 'hex') AS release_key FROM pending_releases
+           WHERE status = ? AND tool = ? AND group_id = ? AND klass = ? LIMIT ?`,
+          [PENDING.waiting, c.tool, c.group_id, c.klass, limit]
+        ))
+        discarded += await this.discardKeys(rows, samplePercent)
+      } while (rows.length === limit && rows.some((r) => !explored(r.release_key, samplePercent)))
+      await db.rawQuery(
+        `UPDATE pending_classes SET swept_at = now() WHERE tool = ? AND group_id = ? AND klass = ?`,
+        [c.tool, c.group_id, c.klass]
+      )
+    }
+    const { rows: signatures } = await db.rawQuery(
+      `SELECT tool, group_id, poster FROM pending_signatures WHERE dropped AND swept_at IS NULL ORDER BY group_id, tool, poster LIMIT ?`,
+      [perSweep]
+    )
+    for (const s of signatures) {
+      let rows: any[]
+      do {
+        // The poster's rows by the signature index (on the poster's md5: a poster can be longer than a btree
+        // key), a tool's and group's by the ladder's: never a test on the rows, which was a scan of the whole
+        // table per signature on the HDD (LOG.md (134)).
+        ;({ rows } =
+          s.poster === ''
+            ? await db.rawQuery(
+                `SELECT encode(release_key, 'hex') AS release_key FROM pending_releases
+                 WHERE status = ? AND group_id = ? AND tool = ? LIMIT ?`,
+                [PENDING.waiting, s.group_id, s.tool, limit]
+              )
+            : await db.rawQuery(
+                `SELECT encode(release_key, 'hex') AS release_key FROM pending_releases
+                 WHERE status = ? AND group_id = ? AND tool = ? AND md5(poster) = md5(?::text) AND poster = ? LIMIT ?`,
+                [PENDING.waiting, s.group_id, s.tool, s.poster, s.poster, limit]
+              ))
+        discarded += await this.discardKeys(rows, samplePercent)
+      } while (rows.length === limit && rows.some((r) => !explored(r.release_key, samplePercent)))
+      await db.rawQuery(
+        `UPDATE pending_signatures SET swept_at = now() WHERE tool = ? AND group_id = ? AND poster = ?`,
+        [s.tool, s.group_id, s.poster]
+      )
+    }
+    return discarded
+  }
+
+  /** The rows discarded (status 2), the explored sample excepted, their classes' waiting counts lowered; how many. */
+  private async discardKeys(rows: any[], samplePercent: number): Promise<number> {
+    const keys = rows
+      .map((r: any) => r.release_key as string)
+      .filter((key: string) => !explored(key, samplePercent))
+      .sort()
+    if (keys.length === 0) return 0
+    // Locked in key order (the record's lock order, LOG.md (137)): a plain `key = ANY(subquery)` locks in scan
+    // order and deadlocked against a batch upsert; ORDER BY sorts before FOR UPDATE locks.
+    const { rows: gone } = await db.rawQuery(
+      `UPDATE pending_releases p SET status = ?
+       FROM (SELECT release_key FROM pending_releases
+             WHERE status = ? AND release_key = ANY(SELECT decode(k, 'hex') FROM unnest(?::text[]) AS k)
+             ORDER BY release_key FOR UPDATE) k
+       WHERE p.release_key = k.release_key
+       RETURNING p.tool, p.group_id, p.klass`,
+      [PENDING.discarded, PENDING.waiting, keys]
+    )
+    const counts = new Map<string, { tool: string; group_id: number; klass: number; n: number }>()
+    for (const g of gone) {
+      const key = `${g.tool}\t${g.group_id}\t${g.klass}`
+      const entry = counts.get(key) ?? {
+        tool: g.tool,
+        group_id: Number(g.group_id),
+        klass: Number(g.klass),
+        n: 0,
+      }
+      entry.n++
+      counts.set(key, entry)
+    }
+    if (counts.size > 0) {
+      await db.rawQuery(
+        `UPDATE pending_classes c SET waiting = greatest(0, c.waiting - v.n)
+         FROM jsonb_to_recordset(?::jsonb) AS v(tool text, group_id int, klass int, n int)
+         WHERE c.tool = v.tool AND c.group_id = v.group_id AND c.klass = v.klass`,
+        [JSON.stringify(lockOrder(counts.values(), (c) => `${c.tool}\t${c.group_id}\t${c.klass}`))]
+      )
+    }
+    return gone.length
+  }
+}
+
+export default new PendingLadder()

app/services/pending_namer.ts

@@ -0,0 +1,253 @@
+import db from '@adonisjs/lucid/services/db'
+import logger from '@adonisjs/core/services/logger'
+import { setTimeout as sleep } from 'node:timers/promises'
+import { promisify } from 'node:util'
+import { unzip } from 'node:zlib'
+import env from '#start/env'
+import coordinatorConfig from '#config/coordinator'
+import { XoverClient } from '#lib/nntp/index'
+import { RangeIndexer, type ReleaseChunk } from '#lib/edge/index'
+import { identityKey } from '#lib/release/keys'
+import BatchStore from '#services/batch_store'
+import ChunkPartitions from '#services/chunk_partitions'
+import Corpus from '#services/corpus'
+import RuleSets from '#services/rule_sets'
+import { PENDING } from '#services/pending_ladder'
+
+const inflate = promisify(unzip)
+
+interface NamedRow {
+  release_key: Buffer
+  identity_key: Buffer
+  group_id: number
+  named: string
+  first_article: number
+  last_article: number
+  file_total: number | null
+  tries: number
+}
+
+/** The namer gives a release up after this many failed moves (its articles gone from the server and the corpus). */
+const MAX_MOVES = 3
+
+/**
+ * The widest article span the namer reads again (a busy group posts 50M articles a day, a large upload among
+ * them spans a million or two): one wider is a name the edge saw again much later ("v-lgpc.3ds"), and its span
+ * of segments would not fit the indexer's memory (LOG.md (133)).
+ */
+const MAX_SPAN = 3_000_000
+
+/**
+ * The move into the catalog (REBUILD.md decision f): a pending release a probe named (status 4) gets its segments
+ * read again, from the corpus when its range is stored, else from the server (XOVER over its article span, one
+ * connection of its own, opened for it and closed after), through the TypeScript indexer, and is written as a
+ * release with the name the probe found (source weak-stem, the name as inner name), its chunk and its probe
+ * candidates, which the prober answers from probe_cache. Runs inside the prober.
+ */
+export class PendingNamer {
+  private client: XoverClient | null = null
+  private usedAt = 0
+
+  constructor(private corpusDir: string | undefined = coordinatorConfig.corpus.dir) {}
+
+  async run(stopped: () => boolean) {
+    while (!stopped()) {
+      let rows: NamedRow[] = []
+      try {
+        rows = await this.take(20)
+        for (const row of rows) {
+          if (stopped()) break
+          try {
+            await this.move(row)
+          } catch (error: any) {
+            logger.warn(
+              `Naming ${row.release_key.toString('hex')} "${row.named}" failed: ${error.message}`
+            )
+            await db.rawQuery(
+              `UPDATE pending_releases SET tries = tries + 1, probed_at = now() WHERE release_key = ?`,
+              [row.release_key]
+            )
+          }
+        }
+      } catch (error: any) {
+        logger.error(`Naming: ${error.message}`)
+      }
+      if (this.client && Date.now() - this.usedAt > coordinatorConfig.nntpIdleCloseMs) {
+        this.client.close()
+        this.client = null
+      }
+      if (rows.length === 0)
+        for (let waited = 0; waited < 30_000 && !stopped(); waited += 1000) await sleep(1000)
+    }
+    this.client?.close()
+    this.client = null
+  }
+
+  /** Named releases awaiting their move, the longest waiting first; those failed MAX_MOVES times stay as they are. */
+  async take(n: number): Promise<NamedRow[]> {
+    const { rows } = await db.rawQuery(
+      `SELECT release_key, identity_key, group_id, named, first_article, last_article, file_total, tries
+       FROM pending_releases WHERE status = ? AND tries < ? ORDER BY probed_at LIMIT ?`,
+      [PENDING.naming, MAX_MOVES, n]
+    )
+    return rows.map((r: any) => ({
+      ...r,
+      group_id: Number(r.group_id),
+      first_article: Number(r.first_article),
+      last_article: Number(r.last_article),
+      tries: Number(r.tries),
+    }))
+  }
+
+  /** One named release into the catalog. */
+  async move(row: NamedRow): Promise<void> {
+    if (row.last_article - row.first_article > MAX_SPAN)
+      throw new Error(
+        `its articles ${row.first_article}–${row.last_article} span more than ${MAX_SPAN}`
+      )
+    const group = await db.from('groups').where('id', row.group_id).firstOrFail()
+    const range = await db
+      .from('ranges')
+      .where('group_id', row.group_id)
+      .where('start_id', '<=', row.first_article)
+      .where('end_id', '>=', row.first_article)
+      .orderBy('id')
+      .first()
+    if (!range) throw new Error(`no range of ${group.name} holds article ${row.first_article}`)
+    const rules = await RuleSets.current()
+    const ix = new RangeIndexer({
+      group: group.name,
+      start: row.first_article,
+      end: row.last_article,
+      rules: { ...rules, samplePercent: 0 },
+      // No file given up before the end: the span is read chunk by chunk (one text of it topped V8's string).
+      idleChunks: Number.MAX_SAFE_INTEGER,
+      edgeArticles: 0,
+      // The edge wrote it partial when its file is short of parts on the server: so must the re-read.
+      partialFiles: true,
+    })
+    for await (const piece of this.overviewPieces(group.name, row))
+      ix.addText(piece.text, piece.next)
+    ix.finish()
+    const batch = ix.take()
+    const found = batch.releases.find((r) => identityKey(r.key).equals(row.identity_key))
+    if (!found)
+      throw new Error(
+        `its articles ${row.first_article}–${row.last_article} of ${group.name} no longer show it`
+      )
+    // The catalog row under the pending key (the identity, the group, the first article the edge saw).
+    const chunk: ReleaseChunk = {
+      ...found,
+      name: row.named,
+      source: 'weak-stem',
+      readable: true,
+      finished: true,
+      edge: false,
+      firstArticle: row.first_article,
+      fileTotal: row.file_total ?? found.fileTotal,
+      tool: undefined,
+    }
+    const candidates = batch.candidates
+      .filter((c) => c.releaseKey === found.key)
+      .map((c) => ({ ...c, upload: found.upload }))
+    // The chunk's month partition, as the batch store makes sure of before its writes.
+    await ChunkPartitions.ensure()
+    await db.transaction(async (trx) => {
+      const blockId = await BatchStore.blockOf(trx, range)
+      await BatchStore.writeReleases(
+        trx,
+        { groupId: row.group_id, blockId },
+        row.first_article,
+        [chunk],
+        candidates
+      )
+      await trx.rawQuery(
+        `UPDATE releases SET inner_name = ?, inspected_at = now() WHERE release_key = ? AND inner_name IS NULL`,
+        [row.named.slice(0, 255), row.release_key]
+      )
+      await trx.rawQuery(
+        `UPDATE pending_releases SET status = ?, probed_at = now() WHERE release_key = ?`,
+        [PENDING.named, row.release_key]
+      )
+    })
+  }
+
+  /**
+   * The overview lines of the release's article span, a chunk of articles at a time with the first article after
+   * it: the stored corpus chunks when its ranges are stored and contiguous, else XOVER in windows of a chunk.
+   */
+  private async *overviewPieces(
+    groupName: string,
+    row: NamedRow
+  ): AsyncGenerator<{ text: string; next: number }> {
+    const size = coordinatorConfig.chunkArticles
+    const stored = await this.storedChunks(groupName, row)
+    if (stored) {
+      for (const { raw, next } of stored)
+        yield { text: (await inflate(raw)).toString('latin1'), next }
+      return
+    }
+    this.client ??= await XoverClient.connect({
+      host: env.get('NNTP_HOST')!,
+      port: env.get('NNTP_PORT', 563),
+      user: env.get('NNTP_USER'),
+      password: env.get('NNTP_PASSWORD'),
+      secure: env.get('NNTP_SECURE', true),
+      compress: true,
+      timeoutMs: coordinatorConfig.nntpTimeoutMs,
+    })
+    this.usedAt = Date.now()
+    try {
+      await this.client.group(groupName)
+      for (let from = row.first_article; from <= row.last_article; from += size) {
+        const to = Math.min(from + size - 1, row.last_article)
+        // One zlib/gzip stream, as the server sent it or deflated by the client (the corpus stores it so).
+        const raw = await this.client.xoverRaw(from, to)
+        yield { text: (await inflate(raw)).toString('latin1'), next: to + 1 }
+      }
+    } catch (error) {
+      this.client.close()
+      this.client = null
+      throw error
+    }
+  }
+
+  /** The span's stored corpus chunks (compressed, with the first article after each), or null when any is missing. */
+  private async storedChunks(
+    groupName: string,
+    row: NamedRow
+  ): Promise<{ raw: Buffer; next: number }[] | null> {
+    if (!this.corpusDir) return null
+    const stored = await db
+      .from('ranges')
+      .where('group_id', row.group_id)
+      .whereNotNull('stored_at')
+      .where('start_id', '<=', row.last_article)
+      .where('end_id', '>=', row.first_article)
+      .orderBy('start_id')
+    if (
+      stored.length === 0 ||
+      Number(stored[0].start_id) > row.first_article ||
+      Number(stored[stored.length - 1].end_id) < row.last_article
+    )
+      return null
+    const size = coordinatorConfig.chunkArticles
+    const chunks: { raw: Buffer; next: number }[] = []
+    for (const r of stored) {
+      const range = { group: groupName, start: Number(r.start_id), end: Number(r.end_id) }
+      for (
+        let from =
+          range.start +
+          Math.floor((Math.max(row.first_article, range.start) - range.start) / size) * size;
+        from <= Math.min(row.last_article, range.end);
+        from += size
+      ) {
+        const to = Math.min(from + size - 1, range.end)
+        const raw = await Corpus.read(range, { from, to }, this.corpusDir)
+        if (!raw) return null
+        chunks.push({ raw, next: to + 1 })
+      }
+    }
+    return chunks
+  }
+}

app/services/planner.ts

@@ -35,20 +35,8 @@ class Planner {
 
   async planGroup(group: { id: number; name: string }): Promise<number> {
     const info = await NntpPool.use((client) => client.group(group.name))
-    const size = coordinatorConfig.rangeArticles
     const last = await db.from('ranges').where('group_id', group.id).max('end_id as end').first()
-    const from = Math.max(info.low, Number(last?.end ?? 0) + 1)
-    const span = Math.max(1, info.high - info.low)
-    const rows = []
-    for (let start = from; start <= info.high; start += size) {
-      rows.push({
-        group_id: group.id,
-        start_id: start,
-        end_id: Math.min(start + size - 1, info.high),
-        next_id: start,
-        priority: Math.floor((1000 * (start - info.low)) / span),
-      })
-    }
+    const rows = historyRows(group.id, info, Number(last?.end ?? 0), coordinatorConfig.rangeArticles)
     for (let i = 0; i < rows.length; i += 2000) await db.table('ranges').multiInsert(rows.slice(i, i + 2000))
     await db.from('groups').where('id', group.id).update({ low: info.low, high: info.high, planned_at: new Date() })
     return rows.length
@@ -75,32 +63,99 @@ class Planner {
    * numbers come from LIST ACTIVE, one command per name prefix (GROUP per
    * group took minutes for 1000 groups).
    */
-  async planTail(now = new Date()): Promise<{ groups: number; ranges: number }> {
+  async planTail(now = new Date()): Promise<TailPlanned> {
     const groups = await db.from('groups').where('active', true).where('tail', true).whereNotNull('planned_at').select('id', 'name', 'planned_at')
-    if (groups.length === 0) return { groups: 0, ranges: 0 }
-    const { rows: ends } = await db.rawQuery(`SELECT DISTINCT ON (group_id) group_id, end_id FROM ranges ORDER BY group_id, start_id DESC`)
-    const plannedTo = new Map<number, number>(ends.map((r: any) => [Number(r.group_id), Number(r.end_id)]))
+    if (groups.length === 0) return { groups: 0, ranges: 0, extended: 0 }
+    const { rows: ends } = await db.rawQuery(
+      `SELECT DISTINCT ON (group_id) group_id, id, start_id, end_id,
+              (status = 'planned' AND priority = ? AND NOT capture AND spooled_chunks = 0 AND attempts = 0 AND node_id IS NULL) AS open
+       FROM ranges ORDER BY group_id, start_id DESC`,
+      [coordinatorConfig.tail.priority]
+    )
+    const last = new Map<number, { id: number; start: number; end: number; open: boolean }>(
+      ends.map((r: any) => [Number(r.group_id), { id: Number(r.id), start: Number(r.start_id), end: Number(r.end_id), open: r.open === true }])
+    )
     const highs = new Map<string, number>()
     await NntpPool.use(async (client) => {
       for (const pattern of new Set(groups.map((g) => wildmatOf(g.name)))) {
         for (const g of await client.listActive(pattern)) highs.set(g.name, g.high)
       }
     })
-    const rows = tailRanges(
-      groups.map((g) => ({ id: Number(g.id), plannedTo: plannedTo.get(Number(g.id)) ?? null, plannedAt: new Date(g.planned_at), high: highs.get(g.name) ?? null })),
+    const plan = tailRanges(
+      groups.map((g) => {
+        const l = last.get(Number(g.id))
+        return { id: Number(g.id), plannedTo: l?.end ?? null, plannedAt: new Date(g.planned_at), high: highs.get(g.name) ?? null, open: l?.open ? { id: l.id, start: l.start } : null }
+      }),
       now,
       { rangeArticles: coordinatorConfig.rangeArticles, ...coordinatorConfig.tail }
     )
-    if (rows.length === 0) return { groups: 0, ranges: 0 }
-    await db.transaction(async (trx) => {
-      await trx.table('ranges').multiInsert(rows)
-      for (const id of new Set(rows.map((r) => r.group_id))) {
-        const high = Math.max(...rows.filter((r) => r.group_id === id).map((r) => r.end_id))
-        await trx.from('groups').where('id', id).update({ high, planned_at: now })
-      }
-    })
-    return { groups: new Set(rows.map((r) => r.group_id)).size, ranges: rows.length }
+    return this.applyTail(plan, now)
   }
+
+  /** The tail's plan written: the open ranges extended (one taken meanwhile gets the slice as a row instead), the rows added. */
+  async applyTail(plan: TailPlan, now = new Date()): Promise<TailPlanned> {
+    const rows = [...plan.rows]
+    let extended = 0
+    if (rows.length === 0 && plan.extended.length === 0) return { groups: 0, ranges: 0, extended: 0 }
+    await db.transaction(async (trx) => {
+      for (const e of plan.extended) {
+        const { rows: done } = await trx.rawQuery(
+          `UPDATE ranges SET end_id = ? WHERE id = ? AND status = 'planned' AND spooled_chunks = 0 AND attempts = 0 AND node_id IS NULL AND end_id = ? RETURNING id`,
+          [e.end_id, e.id, e.from - 1]
+        )
+        if (done.length > 0) extended++
+        else rows.push({ group_id: e.group_id, start_id: e.from, end_id: e.end_id, next_id: e.from, priority: coordinatorConfig.tail.priority })
+      }
+      if (rows.length > 0) await trx.table('ranges').multiInsert(rows)
+      const highs = new Map<number, number>()
+      for (const r of [...rows, ...plan.extended]) highs.set(r.group_id, Math.max(highs.get(r.group_id) ?? 0, r.end_id))
+      for (const [id, high] of highs) await trx.from('groups').where('id', id).update({ high, planned_at: now })
+    })
+    return { groups: new Set([...rows, ...plan.extended].map((r) => r.group_id)).size, ranges: rows.length, extended }
+  }
+
+  /**
+   * Adjacent planned tail ranges of a group and priority, never spooled or taken, merged into ranges of at most
+   * RANGE_ARTICLES (ranges:merge-tail): the tail planned a sliver a group every round while nothing leased the tail
+   * (127k rows of a median 14k articles on 2026-10-27; LOG.md (118)). The first row of a run grows, the others go.
+   */
+  async mergeTail(dryRun = false): Promise<{ candidates: number; merged: number; dropped: number }> {
+    const { rows: candidates } = await db.rawQuery(
+      `SELECT id, group_id, start_id, end_id, priority FROM ranges
+       WHERE status = 'planned' AND priority < 0 AND NOT capture AND spooled_chunks = 0 AND attempts = 0 AND node_id IS NULL
+       ORDER BY group_id, start_id`
+    )
+    const merges = tailMerges(
+      candidates.map((r: any) => ({ id: Number(r.id), group_id: Number(r.group_id), start_id: Number(r.start_id), end_id: Number(r.end_id), priority: Number(r.priority) })),
+      coordinatorConfig.rangeArticles
+    )
+    const dropped = merges.reduce((n, m) => n + m.drop.length, 0)
+    if (dryRun || merges.length === 0) return { candidates: candidates.length, merged: merges.length, dropped }
+    await db.transaction(async (trx) => {
+      // The lease picker reads several tables under this lock; the rows' guards catch a range taken meanwhile.
+      await trx.rawQuery(`SELECT pg_advisory_xact_lock(hashtextextended('group-lease', 0))`)
+      const { rows: gone } = await trx.rawQuery(
+        `DELETE FROM ranges WHERE id = ANY(?::bigint[]) AND status = 'planned' AND spooled_chunks = 0 AND attempts = 0 AND node_id IS NULL RETURNING id`,
+        [merges.flatMap((m) => m.drop)]
+      )
+      if (gone.length !== dropped) throw new Error(`${dropped - gone.length} of the ranges to drop were taken meanwhile; nothing merged`)
+      const { rows: grown } = await trx.rawQuery(
+        `UPDATE ranges r SET end_id = v.end_id FROM (SELECT unnest(?::bigint[]) AS id, unnest(?::bigint[]) AS end_id) v
+         WHERE r.id = v.id AND r.status = 'planned' AND r.spooled_chunks = 0 AND r.attempts = 0 AND r.node_id IS NULL RETURNING r.id`,
+        [merges.map((m) => m.id), merges.map((m) => m.end_id)]
+      )
+      if (grown.length !== merges.length) throw new Error(`${merges.length - grown.length} of the ranges to grow were taken meanwhile; nothing merged`)
+    })
+    return { candidates: candidates.length, merged: merges.length, dropped }
+  }
+}
+
+export interface TailPlanned {
+  groups: number
+  /** Rows added. */
+  ranges: number
+  /** Open ranges grown. */
+  extended: number
 }
 
 /** A LIST ACTIVE pattern for a group: its first two name parts (`alt.binaries.*`), or the name itself. */
@@ -117,28 +172,119 @@ export interface TailGroup {
   plannedAt: Date
   /** The server's newest article (null: not listed). */
   high: number | null
+  /**
+   * Its last range when the tail may grow it instead of adding a row: a planned range at the tail's priority that
+   * nothing spooled, took or tried (its start: the cut at rangeArticles counts from there).
+   */
+  open?: { id: number; start: number } | null
+}
+
+export interface TailRow {
+  group_id: number
+  start_id: number
+  end_id: number
+  next_id: number
+  priority: number
+}
+
+export interface TailPlan {
+  /** New ranges. */
+  rows: TailRow[]
+  /** Open ranges grown to end_id (from: the first new article, the old end + 1). */
+  extended: { id: number; group_id: number; from: number; end_id: number }[]
 }
 
 /**
- * The tail's new ranges: per group, the articles after its last range, in
- * ranges of at most `rangeArticles`, once there are `minArticles` of them or
- * the group's last planning is `maxDelayMinutes` ago.
+ * The tail's new ranges: per group, the articles after its last range, once there are `minArticles` of them or
+ * the group's last planning is `maxDelayMinutes` ago. An open last range grows up to `rangeArticles` first, so a
+ * tail nothing leases stays one range a group instead of a sliver a round; the rest is cut at `rangeArticles`.
  */
 export function tailRanges(
   groups: TailGroup[],
   now: Date,
   options: { rangeArticles: number; minArticles: number; maxDelayMinutes: number; priority: number }
-) {
-  const rows: { group_id: number; start_id: number; end_id: number; next_id: number; priority: number }[] = []
+): TailPlan {
+  const rows: TailRow[] = []
+  const extended: TailPlan['extended'] = []
   for (const g of groups) {
     if (g.plannedTo === null || g.high === null || g.high <= g.plannedTo) continue
     const waiting = g.high - g.plannedTo
     const waited = (now.getTime() - g.plannedAt.getTime()) / 60_000
     if (waiting < options.minArticles && waited < options.maxDelayMinutes) continue
-    for (let start = g.plannedTo + 1; start <= g.high; start += options.rangeArticles) {
+    let start = g.plannedTo + 1
+    if (g.open && g.open.start + options.rangeArticles - 1 >= start) {
+      const end = Math.min(g.open.start + options.rangeArticles - 1, g.high)
+      extended.push({ id: g.open.id, group_id: g.id, from: start, end_id: end })
+      start = end + 1
+    }
+    for (; start <= g.high; start += options.rangeArticles) {
       rows.push({ group_id: g.id, start_id: start, end_id: Math.min(start + options.rangeArticles - 1, g.high), next_id: start, priority: options.priority })
     }
   }
+  return { rows, extended }
+}
+
+export interface TailMerge {
+  /** The run's first range, grown to end_id. */
+  id: number
+  end_id: number
+  /** The ranges it absorbed. */
+  drop: number[]
+}
+
+/**
+ * Runs of adjacent ranges of one group and priority (sorted by group and start) packed into ranges of at most
+ * `rangeArticles`: the first range of each packed run grows, the others are dropped. Only runs that merge anything.
+ */
+export function tailMerges(
+  ranges: { id: number; group_id: number; start_id: number; end_id: number; priority: number }[],
+  rangeArticles: number
+): TailMerge[] {
+  const merges: TailMerge[] = []
+  let cur: (TailMerge & { group_id: number; start_id: number; priority: number }) | null = null
+  const flush = () => {
+    if (cur && cur.drop.length > 0) merges.push({ id: cur.id, end_id: cur.end_id, drop: cur.drop })
+  }
+  for (const r of ranges) {
+    if (cur && r.group_id === cur.group_id && r.priority === cur.priority && r.start_id === cur.end_id + 1 && r.end_id - cur.start_id + 1 <= rangeArticles) {
+      cur.end_id = r.end_id
+      cur.drop.push(r.id)
+      continue
+    }
+    flush()
+    cur = { id: r.id, group_id: r.group_id, start_id: r.start_id, end_id: r.end_id, priority: r.priority, drop: [] }
+  }
+  flush()
+  return merges
+}
+
+/**
+ * The rows a group's history is cut into, from `lastEnd + 1` (0: the group has no ranges yet) up to the high
+ * watermark: RANGE_ARTICLES slices at the position priority. A group planned for the first time also gets one
+ * "ancient" range, articles 1 … low − 1: what had expired on the server before it was planned. Status
+ * 'ancient', never leased or fetched; it keeps the tiling complete from article 1 (migration 6700 added it for
+ * the groups planned before 2026-10-28).
+ */
+export function historyRows(
+  groupId: number,
+  info: { low: number; high: number },
+  lastEnd: number,
+  size: number
+): Record<string, unknown>[] {
+  const rows: Record<string, unknown>[] = []
+  if (lastEnd === 0 && info.low > 1)
+    rows.push({ group_id: groupId, start_id: 1, end_id: info.low - 1, next_id: 1, priority: 0, status: 'ancient' })
+  const from = Math.max(info.low, lastEnd + 1)
+  const span = Math.max(1, info.high - info.low)
+  for (let start = from; start <= info.high; start += size) {
+    rows.push({
+      group_id: groupId,
+      start_id: start,
+      end_id: Math.min(start + size - 1, info.high),
+      next_id: start,
+      priority: Math.floor((1000 * (start - info.low)) / span),
+    })
+  }
   return rows
 }
 

app/services/prefetcher.ts

@@ -425,10 +425,23 @@ export class Prefetcher {
       )
     // Leased ranges, checked against the spool's files whatever spooled_at says: a worker waits for them.
     const leased = await rows(db.query()).where('r.status', 'leased').orderBy('r.leased_at')
+    // The groups workers hold (group_leases; REBUILD.md design 1), by lane: the prefetcher runs ahead of their
+    // cursors (a group's history and tail lanes are two cursors).
+    const held: { group: number; lane: string }[] = (
+      await db
+        .from('group_leases')
+        .whereNotNull('node_id')
+        .where(
+          'heartbeat_at',
+          '>=',
+          new Date(Date.now() - coordinatorConfig.leaseTtlSeconds * 1000)
+        )
+        .select('group_id', 'lane')
+    ).map((l) => ({ group: Number(l.group_id), lane: l.lane }))
     // Planned ranges ahead: those started already, then new ones up to SPOOL_AHEAD_PER_LEASE per lease.
     const ahead = Math.max(
       coordinatorConfig.spool.minAhead,
-      coordinatorConfig.spool.aheadPerLease * leased.length
+      coordinatorConfig.spool.aheadPerLease * Math.max(leased.length, held.length)
     )
     const started = await rows(db.query())
       .where('r.status', 'planned')
@@ -448,14 +461,37 @@ export class Prefetcher {
     const want = Math.max(ahead - Number(ready), short)
     // The catalog at its cap on the NVMe: no history fetched ahead (CatalogSpace).
     const full = await CatalogSpace.full()
+    // Ahead of each held cursor: the lane's next ranges in article order (the history lane's within what the group
+    // leases hand out: LEASE_MIN/MAX_PRIORITY, LEASE_CAPTURE_ONLY, and none at the catalog's cap; the tail lane's
+    // at the tail's priority), a range of every held cursor before any cursor's second; without held groups (no
+    // window worker running) nothing is fetched ahead for workers.
+    const { leasePriorityMin, leasePriorityMax, leaseCaptureOnly } = coordinatorConfig
+    const tailPriority = coordinatorConfig.tail.priority
+    const heldHistory = full ? [] : held.filter((h) => h.lane === 'history').map((h) => h.group)
+    const heldTail = held.filter((h) => h.lane === 'tail').map((h) => h.group)
     const fresh =
-      room && want > 0
+      room && want > 0 && heldHistory.length + heldTail.length > 0
         ? await rows(db.query())
             .where('r.status', 'planned')
             .where('r.spooled_chunks', 0)
             .whereRaw('r.next_id <= r.end_id')
-            .where((q: any) => (full ? q.where('r.priority', '<', 0) : q))
-            .orderBy([{ column: 'r.priority' }, { column: 'r.id' }])
+            .where((q: any) =>
+              q
+                .where((h: any) =>
+                  h
+                    .whereIn('r.group_id', heldHistory)
+                    .where('r.priority', '>', tailPriority)
+                    .whereBetween('r.priority', [leasePriorityMin, leasePriorityMax])
+                    .where((c: any) => (leaseCaptureOnly ? c.where('r.capture', true) : c))
+                )
+                .orWhere((t: any) =>
+                  t.whereIn('r.group_id', heldTail).where('r.priority', '<=', tailPriority)
+                )
+            )
+            .orderByRaw(
+              'row_number() OVER (PARTITION BY r.group_id, r.priority <= ? ORDER BY r.start_id), r.priority, r.start_id',
+              [tailPriority]
+            )
             .limit(want)
         : []
     const targets: Target[] = []
@@ -464,9 +500,13 @@ export class Prefetcher {
       const end = Number(row.end_id)
       const total = Spool.chunksOf({ start, end }).length
       const missing: ChunkSpan[] = []
+      // A chunk already in the corpus isn't spooled: the stream serves it from there (REBUILD.md design 1).
+      const corpusRange = row.capture === true ? { group: row.group as string, start, end } : null
       for (const span of Spool.chunksOf({ start, end }, Number(row.next_id))) {
-        if (!this.claimed.has(`${row.id}:${span.from}`) && !(await Spool.has(Number(row.id), span)))
-          missing.push(span)
+        if (this.claimed.has(`${row.id}:${span.from}`) || (await Spool.has(Number(row.id), span)))
+          continue
+        if (corpusRange && (await Corpus.has(corpusRange, span))) continue
+        missing.push(span)
       }
       if (missing.length > 0) {
         targets.push({
@@ -506,7 +546,10 @@ export class Prefetcher {
    * Up to corpus.ahead capture ranges whose chunks aren't all in the corpus, by priority: planned ones not started
    * (not leased soon), and done ones (a window of every group, ranges:capture, takes those indexed already too);
    * as many of a group as its connections allow (perGroup / perRange): in id order the 16 were one group's, which
-   * took 4 of the loops and left the rest idle (30k articles/s; LOG.md 2026-10-23 (100)). A range found complete
+   * took 4 of the loops and left the rest idle (30k articles/s; LOG.md 2026-10-23 (100)). The candidates are taken
+   * a range of every group at a time (the n-th unstored range of each group before any group's n+1-th): a group's
+   * ranges have consecutive ids, and once the small groups of a window were stored the 512 lowest ids were all one
+   * group's, two targets on two connections (10 ranges an hour; LOG.md 2026-10-25 (110)). A range found complete
    * gets stored_at and isn't read again.
    */
   private async corpusTargets(rows: (q: any) => any, taken: Set<number>): Promise<Target[]> {
@@ -518,13 +561,15 @@ export class Prefetcher {
     const groups = new Map<string, number>()
     const candidates = await rows(db.query())
       .where('r.capture', true)
+      // A group taken off the list (groups.active) isn't worth the connections.
+      .where('g.active', true)
       .whereNull('r.stored_at')
       .where((q: any) =>
         q
           .where((p: any) => p.where('r.status', 'planned').where('r.spooled_chunks', 0))
           .orWhere('r.status', 'done')
       )
-      .orderBy([{ column: 'r.priority' }, { column: 'r.id' }])
+      .orderByRaw('r.priority, row_number() OVER (PARTITION BY r.group_id ORDER BY r.id), r.id')
       .limit(ahead * 16)
     const targets: Target[] = []
     for (const row of candidates) {

app/services/prober.ts

@@ -9,12 +9,17 @@ import { inspect, innerName, PROBE_VERSION, type Inspection } from '#lib/probe/i
 import { NntpShare } from '#services/nntp_shares'
 import { DiskSpace } from '#services/disk_space'
 import { decidePar2Completeness } from '#services/par2_completeness'
+import { twinKey, twinNames } from '#services/hash_twins'
+import PendingLadder, { type PendingProbe } from '#services/pending_ladder'
+import { PendingNamer } from '#services/pending_namer'
 
 interface Candidate {
   id: number
   releaseId: number
   kind: string
   messageId: string
+  /** A pending release's target (the pending ladder), not a catalog candidate: id and releaseId are 0. */
+  pending?: PendingProbe
 }
 
 type Outcome = { candidate: Candidate; status: 'done' | 'missing' | 'error' | 'waiting'; inspection?: Inspection; cached?: boolean }
@@ -34,6 +39,10 @@ const TIERS: { kind: string; unreadable: boolean }[] = [
   { kind: '7z-tail', unreadable: false },
 ]
 const FLUSH_MS = 2000
+/** Between sweeps that discard the waiting releases of dropped signatures. */
+const SWEEP_MS = 5 * 60_000
+/** A flush that deadlocks (against the batch store's pending upsert) keeps its outcomes for the next, this often. */
+const DEADLOCK_RETRIES = 5
 
 /**
  * JSON for a jsonb parameter, without NUL characters: jsonb can't hold \u0000 (a name read from an article had one),
@@ -43,6 +52,20 @@ function jsonb(value: unknown): string {
   return JSON.stringify(value, (_key, v) => (typeof v === 'string' ? v.replaceAll('\0', '') : v))
 }
 
+/** Rounds of every lane's pipeline still queued when the next claim starts. */
+const LOW_ROUNDS = 8
+
+/**
+ * The candidate queue's levels: the next claim starts below `low`, LOW_ROUNDS rounds of every lane's pipeline (at
+ * ~0.2 s an article and a pipeline of 8, ~13 s of work), and fills it to `target`, twice that. A claim takes 1.5–4 s
+ * and more (random slices of an 11 M-row table); refilling at half of four rounds left 3–6 s, and at 20–24 lanes the
+ * lanes waited for candidates up to 19% of their time (LOG.md 2026-10-28 (120)). At least `minBatch` either way.
+ */
+export function queueLevels(lanes: number, pipeline: number, minBatch: number): { low: number; target: number } {
+  const low = Math.max(minBatch, lanes * pipeline * LOW_ROUNDS)
+  return { low, target: 2 * low }
+}
+
 /** Slices a claim takes from random points of the queue. */
 const SLICES = 20
 const REPORT_MS = 60_000
@@ -61,6 +84,8 @@ export class Prober {
   private refilling: Promise<void> | null = null
   private stopped = false
   private stats = Prober.freshStats()
+  /** Flushes deadlocked in a row, their outcomes kept. */
+  private deadlocks = 0
   /** Lanes running (each one connection, once it has work); at most the share's grant. Counted apart from the set: a lane leaving takes itself off at once. */
   private lanes = new Set<Promise<void>>()
   private running = 0
@@ -72,13 +97,29 @@ export class Prober {
 
   /** Per minute: what was read, and where the lanes' time went (ms summed over lanes; flushes ms on their own). */
   private static freshStats() {
-    return { articles: 0, bytes: 0, done: 0, missing: 0, error: 0, cached: 0, waitMs: 0, idleMs: 0, fetchMs: 0, inspectMs: 0, flushMs: 0, claimMs: 0, since: Date.now() }
+    return { articles: 0, bytes: 0, done: 0, missing: 0, error: 0, cached: 0, pending: 0, discarded: 0, waitMs: 0, idleMs: 0, fetchMs: 0, inspectMs: 0, flushMs: 0, claimMs: 0, since: Date.now() }
   }
+  /** The pending ladder's namer: named pending releases into the catalog, alongside the lanes. */
+  private namer = new PendingNamer()
 
   async run() {
     // A previous run's claims: one prober at a time.
     const { rowCount } = await db.rawQuery(`UPDATE probe_candidates SET status = 'waiting' WHERE status = 'taken'`)
     if (rowCount) logger.info(`Probing: ${rowCount} candidates from a previous run back in the queue.`)
+    const released = await PendingLadder.release()
+    if (released) logger.info(`Probing: ${released} pending releases from a previous run back on the ladder.`)
+    const namer = this.namer.run(() => this.stopped).catch((error) => logger.error(`Naming failed: ${error.message}`))
+    // The ladder's discard sweep beside the lanes: a dropped signature's rows take minutes on the HDD (LOG.md (133)).
+    const sweeper = (async () => {
+      while (!this.stopped) {
+        try {
+          this.stats.discarded += await PendingLadder.discard()
+        } catch (error: any) {
+          logger.error(`Probing: the discard sweep failed: ${error.message}`)
+        }
+        for (let waited = 0; waited < SWEEP_MS && !this.stopped; waited += 1000) await sleep(1000)
+      }
+    })()
     logger.info(`Probing with ${this.share.min}–${this.share.max} connections (its share of the account's).`)
     const flusher = (async () => {
       while (!this.stopped || this.outcomes.length > 0) {
@@ -108,6 +149,8 @@ export class Prober {
       for (let waited = 0; waited < coordinatorConfig.nntpShares.everyMs && !this.stopped; waited += 1000) await sleep(1000)
     }
     await Promise.all(this.lanes)
+    await namer
+    await sweeper
     await this.share.leave().catch(() => {})
     // What was claimed and not probed goes back.
     for (const candidate of this.queue.splice(0)) this.outcomes.push({ candidate, status: 'waiting' })
@@ -196,13 +239,18 @@ export class Prober {
     }
   }
 
-  /** Candidates claimed at a time: a few rounds of every lane's pipeline (a claim takes seconds; 400 lasted ~2 s at 37 lanes). */
+  /** The queue's low-water mark and the level a claim fills it to (queueLevels). */
+  private get levels() {
+    return queueLevels(this.running, coordinatorConfig.probe.pipeline, coordinatorConfig.probe.batch)
+  }
+
+  /** Claims to the queue's target; the claim after starts when it's below its low-water mark (queueLevels). */
   private get batch() {
-    return Math.max(coordinatorConfig.probe.batch, this.running * coordinatorConfig.probe.pipeline * 4)
+    return this.levels.target
   }
 
   private async next(): Promise<Candidate | undefined> {
-    if (this.queue.length < this.batch / 2) {
+    if (this.queue.length < this.levels.low) {
       this.refilling ??= this.claim().finally(() => (this.refilling = null))
       if (this.queue.length === 0) await this.refilling
     }
@@ -227,6 +275,33 @@ export class Prober {
   private async claimTiers() {
     // The queue empties, the lanes go dry and the share shrinks to its minimum until there's room again.
     if (await this.space.short()) return
+    await this.claimCatalog()
+    await this.claimPending()
+  }
+
+  /** The pending ladder's share of a claim: at least pendingShare of the batch, and whatever the catalog's tiers left. */
+  private async claimPending() {
+    const want = Math.max(Math.floor(this.batch * coordinatorConfig.probe.pendingShare), this.batch - this.queue.length)
+    const probes = await PendingLadder.claim(want)
+    if (probes.length === 0) return
+    const { rows: cached } = await db.rawQuery(`SELECT message_id, status, result FROM probe_cache WHERE message_id = ANY(?) AND version = ?`, [
+      probes.map((p) => p.messageId),
+      PROBE_VERSION,
+    ])
+    const known = new Map(cached.map((c: any) => [c.message_id, c]))
+    for (const probe of probes) {
+      const candidate: Candidate = { id: 0, releaseId: 0, kind: probe.kind, messageId: probe.messageId, pending: probe }
+      const hit: any = known.get(probe.messageId)
+      if (hit) {
+        this.stats.cached++
+        this.outcomes.push({ candidate, status: hit.status, inspection: hit.result ?? undefined, cached: true })
+      } else {
+        this.queue.push(candidate)
+      }
+    }
+  }
+
+  private async claimCatalog() {
     for (const { kind, unreadable } of TIERS) {
       const want = this.batch - this.queue.length
       if (want <= 0) return
@@ -275,20 +350,46 @@ export class Prober {
     }
   }
 
-  /** The outcomes so far: candidates' status, inspections, and the releases' passworded and inner name. */
+  /**
+   * The outcomes so far: candidates' status, inspections, and the releases' passworded and inner name. A deadlock
+   * (the batch store's pending upsert locks the same rows; LOG.md (134)) keeps the outcomes for the next flush
+   * instead of losing the probes.
+   */
   private async flush() {
-    const outcomes = this.outcomes.splice(0)
-    if (outcomes.length === 0) return
-    for (const o of outcomes) if (o.status !== 'waiting') this.stats[o.status]++
+    const taken = this.outcomes.splice(0)
+    if (taken.length === 0) return
+    try {
+      await this.store(taken)
+    } catch (error: any) {
+      if (error?.code === '40P01' && this.deadlocks < DEADLOCK_RETRIES) {
+        this.deadlocks++
+        this.outcomes.unshift(...taken)
+        logger.warn(`Probing: storing results deadlocked, kept for the next flush (${this.deadlocks}).`)
+        return
+      }
+      throw error
+    }
+    this.deadlocks = 0
+    for (const o of taken) if (o.status !== 'waiting') this.stats[o.status]++
+    this.stats.pending += taken.filter((o) => o.candidate.pending && o.status !== 'waiting').length
+  }
+
+  private async store(taken: Outcome[]) {
+    // The ladder's outcomes apart from the catalog's: the next target, a name for the namer, or the end of the climb.
+    const pending = taken.filter((o) => o.candidate.pending)
+    const outcomes = taken.filter((o) => !o.candidate.pending)
     await db.transaction(async (trx) => {
-      await trx.rawQuery(
-        `UPDATE probe_candidates pc SET status = v.status FROM jsonb_to_recordset(?::jsonb) AS v(id bigint, status text) WHERE pc.id = v.id`,
-        [jsonb(outcomes.map((o) => ({ id: o.candidate.id, status: o.status })))]
-      )
+      if (outcomes.length > 0) {
+        await trx.rawQuery(
+          `UPDATE probe_candidates pc SET status = v.status FROM jsonb_to_recordset(?::jsonb) AS v(id bigint, status text) WHERE pc.id = v.id`,
+          [jsonb(outcomes.map((o) => ({ id: o.candidate.id, status: o.status })))]
+        )
+      }
+      await PendingLadder.record(pending.map((o) => ({ probe: o.candidate.pending!, status: o.status, inspection: o.inspection })), trx)
       // Final answers read from the news server, kept by message-id for the next catalog (a reset).
       // One per message-id: the same article is a candidate of more than one release, and ON CONFLICT can't take a
       // row twice in one statement.
-      const fresh = [...new Map(outcomes.filter((o) => !o.cached && o.status !== 'waiting').map((o) => [o.candidate.messageId, o])).values()]
+      const fresh = [...new Map(taken.filter((o) => !o.cached && o.status !== 'waiting').map((o) => [o.candidate.messageId, o])).values()]
       if (fresh.length > 0) {
         await trx.rawQuery(
           `INSERT INTO probe_cache (message_id, status, result, version)
@@ -322,10 +423,20 @@ export class Prober {
         entry.all.push({ kind: r.kind, result: r.result })
         byRelease.set(id, entry)
       }
+      // A twin by content names a release its own inspections didn't (hash_twins.ts; LOG.md (130)).
+      const twins = await twinNames(
+        done.filter((o) => o.inspection!.hash16k && o.inspection!.size).map((o) => ({ hash16k: o.inspection!.hash16k!, size: o.inspection!.size })),
+        trx
+      )
+      const twinOf = new Map<number, string>()
+      for (const o of done) {
+        const twin = o.inspection!.hash16k ? twins.get(twinKey(o.inspection!.hash16k, o.inspection!.size)) : undefined
+        if (twin && twin !== byRelease.get(o.candidate.releaseId)?.name) twinOf.set(o.candidate.releaseId, twin)
+      }
       const updates = [...byRelease].map(([id, { name, all }]) => ({
         id,
         passworded: all.some((i) => i.result.passworded === true),
-        inner_name: innerName(all, name),
+        inner_name: innerName(all, name) ?? twinOf.get(id) ?? null,
       }))
       await trx.rawQuery(
         `UPDATE releases r SET passworded = v.passworded, inner_name = v.inner_name, inspected_at = now()
@@ -365,7 +476,8 @@ export class Prober {
     const { articles, bytes, done, missing, error } = this.stats
     logger.info(
       `Probed ${articles} articles in ${seconds.toFixed(0)} s: ${(articles / seconds).toFixed(1)}/s, ${((bytes * 8) / seconds / 1e6).toFixed(0)} Mbit/s; ` +
-        `${done} inspected, ${missing} missing, ${error} not yEnc; ${this.stats.cached} from the probe cache. ` +
+        `${done} inspected, ${missing} missing, ${error} not yEnc; ${this.stats.cached} from the probe cache; ` +
+        `the pending ladder ${this.stats.pending} probed, ${this.stats.discarded} discarded by signature. ` +
         `Lanes: fetching ${(this.stats.fetchMs / 1000).toFixed(0)} s, waiting for candidates ${(this.stats.waitMs / 1000).toFixed(0)} s, ` +
         `idle ${(this.stats.idleMs / 1000).toFixed(0)} s, inspecting ${(this.stats.inspectMs / 1000).toFixed(1)} s; ` +
         `claims ${(this.stats.claimMs / 1000).toFixed(1)} s, flushes ${(this.stats.flushMs / 1000).toFixed(1)} s.`

app/services/range_blocks.ts

@@ -1,31 +0,0 @@
-import db from '@adonisjs/lucid/services/db'
-
-/**
- * range_blocks (migration 3900): per group, each run of consecutive done ranges with its posting time span, read by
- * release_final(). A few thousand rows from ~650k ranges, rebuilt whole in one statement each time; a run breaks at a
- * range that isn't done or doesn't start where the one before ended.
- */
-class RangeBlocks {
-  async rebuild(): Promise<number> {
-    return db.transaction(async (trx) => {
-      await trx.rawQuery(`DELETE FROM range_blocks`)
-      const result = await trx.rawQuery(
-        `WITH r AS (
-           SELECT group_id, start_id, end_id, status = 'done' AS done, posted_low, posted_high,
-                  min(start_id) OVER (PARTITION BY group_id) AS group_first,
-                  lag(status = 'done') OVER w AS done_before, lag(end_id) OVER w AS end_before
-           FROM ranges WINDOW w AS (PARTITION BY group_id ORDER BY start_id)),
-         runs AS (
-           SELECT *, sum(CASE WHEN done IS DISTINCT FROM done_before OR end_before IS DISTINCT FROM start_id - 1 THEN 1 ELSE 0 END)
-                       OVER (PARTITION BY group_id ORDER BY start_id) AS run
-           FROM r)
-         INSERT INTO range_blocks (group_id, start_id, end_id, posted_low, posted_high, at_group_start)
-         SELECT group_id, min(start_id), max(end_id), min(posted_low), max(posted_high), min(start_id) = min(group_first)
-         FROM runs WHERE done GROUP BY group_id, run`
-      )
-      return result.rowCount ?? 0
-    })
-  }
-}
-
-export default new RangeBlocks()

app/services/release_judge.ts

@@ -1,5 +1,7 @@
 import db from '@adonisjs/lucid/services/db'
 import logger from '@adonisjs/core/services/logger'
+import { SOURCES } from '#lib/release/source'
+import { TINY_MIN_BYTES, NONSENSE_MIN_BYTES } from '#lib/rules/index'
 
 /** Raised when a rule changes: verdicts of an older version are judged again (a full round, right away). */
 export const JUDGE_VERSION = 1
@@ -27,6 +29,14 @@ const MEDIA =
 const SCRAPED_POSTERS = `celeb[-_ ]{0,1}movie|celebwow|celebfx|iluvcelebs|nude[-_ ]{0,1}celebs|celeb-sex-tapes`
 /** An executable's name. */
 const EXECUTABLE = `\\.(exe|scr|com|bat|cmd|msi|lnk|vbs|js|jar|pif|cpl|hta|ps1|dll)$`
+/**
+ * The tiny rule's name tests, lib/rules/rule_set.ts's RELEASE_LIKE_SPACED, EBOOK, TINY_CONTENT and
+ * STUB_EXEMPT_GROUPS in Postgres's dialect (and no "?", see MEDIA).
+ */
+const TINY_RELEASE_LIKE = `(\\y(19|20)[0-9][0-9]\\y|[._ -](720p|1080p|2160p|x264|x265|hevc|bluray|web-{0,1}dl|webrip|hdtv|xvid|s[0-9][0-9]e[0-9][0-9]|german|multi|crack|repack|proper|flac|mp3|xxx|mp4|retail|keygen|portable|incl|v[0-9]+(\\.[0-9]+)+)\\y|-[a-z0-9]{2,}$|\\.[a-z0-9]+\\.[a-z0-9]+\\.)`
+const TINY_EBOOK = `\\y(epub|mobi|azw3{0,1}|pdf|djvu|cbr|cbz|e-{0,1}books{0,1})\\y`
+const TINY_CONTENT = `\\y(subs{0,1}|subtitles{0,1}|subpack|srt|untertitel|ondertitels{0,1}|ondertiteling|legendas{0,1}|napisy|sottotitoli|cracks{0,1}|cracked|keygen|keymaker|trainer|patch|patched|serials{0,1}|activator|loader|nocd|no-cd|fix|covers{0,1}|artwork|booklet|scans{0,1}|proof|sample|info|readme)\\y`
+const TINY_EXEMPT_GROUPS = `(sound|mp3|karaoke|music|audio|flac|lossless|e-{0,1}book|epub|comic|guitar|font|magazine)`
 
 export interface Rule {
   name: string
@@ -88,10 +98,10 @@ export const RULES: Rule[] = [
     // regex on all of releases was 45 s, every round (LOG.md 2026-10-19 (80)).
     sql: (scoped) =>
       scoped
-        ? `SELECT r.id FROM releases r ${only(scoped, 'r.id')}
-          WHERE r.poster ~* '${SCRAPED_POSTERS}' AND r.source = 'subject' AND r.file_total IS NULL`
-        : `SELECT r.id FROM posters p JOIN releases r ON left(r.poster, 200) = p.poster
-          WHERE p.poster ~* '${SCRAPED_POSTERS}' AND r.poster ~* '${SCRAPED_POSTERS}' AND r.source = 'subject' AND r.file_total IS NULL`,
+        ? `SELECT r.id FROM releases r ${only(scoped, 'r.id')} JOIN posters p ON p.id = r.poster_id
+          WHERE p.poster ~* '${SCRAPED_POSTERS}' AND r.source = ${SOURCES.subject} AND r.file_total IS NULL`
+        : `SELECT r.id FROM posters p JOIN releases r ON r.poster_id = p.id
+          WHERE p.poster ~* '${SCRAPED_POSTERS}' AND r.source = ${SOURCES.subject} AND r.file_total IS NULL`,
   },
   {
     name: 'locked',
@@ -99,13 +109,27 @@ export const RULES: Rule[] = [
     // releases_passworded_idx: a scan of all of releases was 30 s a round.
     sql: (scoped) => `SELECT r.id FROM releases r ${only(scoped, 'r.id')} WHERE r.passworded AND NOT EXISTS (SELECT 1 FROM release_passwords p WHERE p.release_id = r.id)`,
   },
+  {
+    name: 'tiny',
+    says: 'one file of one article, 2–100 kB, under a release-like name: the exe-stub flood in miniature',
+    // The catalog's side of the edge's tiny-archive rule (lib/rules/rule_set.ts isTinyArchive; LOG.md (131),
+    // (135)): the same name tests in Postgres's regex dialect ("\y" for "\b", "{0,1}" for "?"), the readable flag
+    // for its scrambled and random tests, the group set for its groups. The 2015–17 flood that the edge indexed
+    // before the rule: 1.5M rows, 98% of them with an archive candidate, the rest lone par2 files and scrambled
+    // names, so no join on the candidates. A scan of releases (9 GB) a full round.
+    sql: (scoped) => `SELECT r.id FROM releases r ${only(scoped, 'r.id')}
+          WHERE r.files = 1 AND r.segments = 1 AND r.bytes >= ${TINY_MIN_BYTES} AND r.bytes < ${NONSENSE_MIN_BYTES}
+            AND r.readable AND NOT coalesce(r.sidecar_only, false)
+            AND r.name ~* '${TINY_RELEASE_LIKE}' AND r.name !~* '${TINY_EBOOK}' AND r.name !~* '${TINY_CONTENT}'
+            AND NOT EXISTS (SELECT 1 FROM group_sets gs WHERE gs.id = r.group_set_id AND gs.groups::text ~* '${TINY_EXEMPT_GROUPS}')`,
+  },
   {
     name: 'unusable',
     says: 'random names, no name inside, no par2 or inner-name merge, no spot, a day after settling, and final',
-    // Final (release_final): next to an unread range its par2 or the rest of its files may still come (LOG.md 2026-10-16 (67)).
+    // Finished (settled_at): a provisional tail release's par2 or the rest of its files may still come.
     sql: (scoped) => `SELECT r.id FROM releases r ${only(scoped, 'r.id')}
           WHERE r.settled_at < now() - interval '1 day' AND NOT r.readable AND r.inner_name IS NULL
-            AND release_final(r.group_id, r.first_article, r.last_article, r.posted_at, r.last_posted_at, r.complete)
+            AND r.settled_at IS NOT NULL
             AND NOT EXISTS (SELECT 1 FROM release_merges m WHERE m.into_id = r.id)
             AND NOT EXISTS (SELECT 1 FROM release_spots s WHERE s.release_id = r.id)
             AND NOT EXISTS (SELECT 1 FROM par2_sets p WHERE p.release_id = r.id)`,

app/services/search_service.ts

@@ -1,6 +1,7 @@
 import db from '@adonisjs/lucid/services/db'
 import logger from '@adonisjs/core/services/logger'
 import NzbService from '#services/nzb_service'
+import { SOURCES, sourceName } from '#lib/release/source'
 
 export type SortOrder = 'newest' | 'oldest' | 'largest' | 'files' | 'name'
 
@@ -51,7 +52,7 @@ export interface ReleaseSummary {
   size: number
   readable: boolean
   complete: boolean | null
-  /** The complete verdict is definite: done ranges cover the release's posting times and 6 hours either side (release_final). */
+  /** The complete verdict is definite: the release is finished (voyager 3: the window closed on it; a tail release is provisional until then). */
   final: boolean
   category: number
   settled: boolean
@@ -84,6 +85,19 @@ function contains(word: string): string {
   return `%${word.replace(/[\\%_]/g, (c) => `\\${c}`)}%`
 }
 
+/**
+ * A search word's condition on release_texts: anywhere in a text (ILIKE), or with a leading "^" at a word's start, the
+ * text's or after a character that's no letter or digit ("^zed" finds "Zed.Dead", "The_Zed", not "Facialized"). Both
+ * use the trigram index.
+ */
+function wordMatch(word: string): { sql: string; binding: string } {
+  if (word.startsWith('^') && word.length > 1) {
+    const rest = word.slice(1).replace(/[\\^$.|?*+()[\]{}]/g, (c) => `\\${c}`)
+    return { sql: 't.text ~* ?', binding: `(^|[^[:alnum:]])${rest}` }
+  }
+  return { sql: 't.text ILIKE ?', binding: contains(word) }
+}
+
 /** Searching and browsing the catalog, for the web interface and Newznab. */
 class SearchService {
   async search(params: SearchParams) {
@@ -91,27 +105,34 @@ class SearchService {
     // Judged junk (app/services/release_judge.ts) is left out unless asked for.
     if (!params.judged) where.push('NOT EXISTS (SELECT 1 FROM release_verdicts v WHERE v.release_id = r.id)')
     const bindings: unknown[] = []
-    // Every word in the name or the inner name the prober found, through their trigram indexes; "The.Block.AU" and "the block au" alike.
+    // Every word in one of the release's texts (its name, the names the prober found inside; release_texts, by its
+    // trigram index, which yields ids: never an OR across the two tables); "The.Block.AU" and "the block au" alike;
+    // "^zed" only at a word's start.
     for (const word of (params.q ?? '').split(/[\s._-]+/).filter(Boolean).slice(0, 8)) {
-      where.push('(r.name ILIKE ? OR r.inner_name ILIKE ?)')
-      bindings.push(contains(word), contains(word))
+      const match = wordMatch(word)
+      where.push(`r.id IN (SELECT t.release_id FROM release_texts t WHERE ${match.sql})`)
+      bindings.push(match.binding)
     }
     if (params.readable) where.push('(r.readable OR r.inner_name IS NOT NULL)')
     if (params.complete) where.push('r.complete')
-    // Incomplete only when final: a release next to unread ranges may yet get its missing files (release_final).
-    if (params.notIncomplete) where.push('(r.complete IS NOT FALSE OR NOT release_final(r.group_id, r.first_article, r.last_article, r.posted_at, r.last_posted_at, r.complete))')
+    // Incomplete only when final: a provisional tail release may yet get its missing files.
+    if (params.notIncomplete) where.push('(r.complete IS NOT FALSE OR r.settled_at IS NULL)')
     if (params.unlocked) where.push('r.passworded IS NOT TRUE')
     if (params.source) {
-      where.push('r.source = ?')
-      bindings.push(params.source)
+      const code = (SOURCES as Record<string, number>)[params.source]
+      if (code === undefined) where.push('false')
+      else {
+        where.push('r.source = ?')
+        bindings.push(code)
+      }
     }
     // A poster search: the matching posters first (posters, by trigram), each one's releases then by
     // releases_poster_newest_idx (below). "poster ILIKE" on releases walked all 50M rows newest first: 30+ min for a
     // poster with few releases; as a filter on the posters found the planner still walked them (LOG.md 2026-10-16 (70)).
-    let posters: string[] | null = null
+    let posters: number[] | null = null
     if (params.poster) {
-      const { rows: matched } = await db.rawQuery(`SELECT poster FROM posters WHERE poster ILIKE ? LIMIT ${MAX_POSTERS}`, [contains(params.poster)])
-      posters = matched.map((m: { poster: string }) => m.poster)
+      const { rows: matched } = await db.rawQuery(`SELECT id FROM posters WHERE poster ILIKE ? LIMIT ${MAX_POSTERS}`, [contains(params.poster)])
+      posters = matched.map((m: { id: string }) => Number(m.id))
     }
     if (params.categories?.length) {
       where.push('r.category = ANY(?)')
@@ -143,14 +164,14 @@ class SearchService {
     const whereSql = where.join(' AND ')
     const countLimit = offset + (params.countLimit ?? COUNT_LIMIT)
     if (posters !== null && posters.length === 0) return { page, perPage, offset, total: 0, totalCapped: false, releases: [] as ReleaseSummary[] }
-    // Per poster found (by first 200 characters, as posters and the index keep them; a btree entry holds at most 2,704
-    // bytes): its releases matching everything else, the first offset + perPage in the order asked; then merged.
+    // Per poster found: its releases matching everything else (releases_poster_newest_idx), the first offset + perPage
+    // in the order asked; then merged.
     const source =
       posters === null
         ? { from: 'releases r', where: whereSql, bindings: [] as unknown[] }
         : {
-            from: `unnest(?::text[]) AS p(poster) CROSS JOIN LATERAL (
-                     SELECT r.* FROM releases r WHERE left(r.poster, 200) = p.poster AND ${whereSql} ORDER BY ${order} LIMIT ${offset + perPage}) r`,
+            from: `unnest(?::int[]) AS p(id) CROSS JOIN LATERAL (
+                     SELECT r.* FROM releases r WHERE r.poster_id = p.id AND ${whereSql} ORDER BY ${order} LIMIT ${offset + perPage}) r`,
             where: 'true',
             bindings: [posters] as unknown[],
           }
@@ -158,16 +179,17 @@ class SearchService {
       posters === null
         ? db.rawQuery(`SELECT count(*) AS total FROM (SELECT 1 FROM releases r WHERE ${whereSql} LIMIT ${countLimit + 1}) x`, bindings)
         : db.rawQuery(
-            `SELECT count(*) AS total FROM (SELECT 1 FROM unnest(?::text[]) AS p(poster) CROSS JOIN LATERAL (
-               SELECT 1 FROM releases r WHERE left(r.poster, 200) = p.poster AND ${whereSql} LIMIT ${countLimit + 1}) x LIMIT ${countLimit + 1}) y`,
+            `SELECT count(*) AS total FROM (SELECT 1 FROM unnest(?::int[]) AS p(id) CROSS JOIN LATERAL (
+               SELECT 1 FROM releases r WHERE r.poster_id = p.id AND ${whereSql} LIMIT ${countLimit + 1}) x LIMIT ${countLimit + 1}) y`,
             [posters, ...bindings]
           )
     const [{ rows }, { rows: counted }] = await Promise.all([
       db.rawQuery(
-        `SELECT r.id, r.name, r.source, r.poster, r.groups, r.posted_at, r.file_total, r.files, r.bytes, r.readable, r.complete, r.category, r.settled_at, r.passworded, r.inner_name,
+        `SELECT r.id, r.name, r.source, po.poster, gs.groups, r.posted_at, r.file_total, r.files, r.bytes, r.readable, r.complete, r.category, r.settled_at, r.passworded, r.inner_name,
                 EXISTS (SELECT 1 FROM release_passwords p WHERE p.release_id = r.id) AS password_known,
-                release_final(r.group_id, r.first_article, r.last_article, r.posted_at, r.last_posted_at, r.complete) AS final
-         FROM ${source.from} WHERE ${source.where} ORDER BY ${order} LIMIT ? OFFSET ?`,
+                r.settled_at IS NOT NULL AS final
+         FROM ${source.from} LEFT JOIN posters po ON po.id = r.poster_id LEFT JOIN group_sets gs ON gs.id = r.group_set_id
+         WHERE ${source.where} ORDER BY ${order} LIMIT ? OFFSET ?`,
         [...source.bindings, ...bindings, perPage, offset]
       ),
       counting,
@@ -182,9 +204,9 @@ class SearchService {
       releases: (rows as any[]).map((r): ReleaseSummary => ({
         id: Number(r.id),
         name: r.name,
-        source: r.source,
-        poster: r.poster,
-        groups: r.groups,
+        source: sourceName(r.source),
+        poster: r.poster ?? null,
+        groups: r.groups ?? [],
         postedAt: r.posted_at,
         fileTotal: r.file_total,
         fileCount: Number(r.files),
@@ -203,20 +225,24 @@ class SearchService {
 
   /** A release with its files, read from its chunks. */
   async release(id: number) {
-    const r = await db.from('releases').where('id', id).first()
+    const {
+      rows: [r],
+    } = await db.rawQuery(
+      `SELECT r.*, po.poster, gs.groups FROM releases r LEFT JOIN posters po ON po.id = r.poster_id LEFT JOIN group_sets gs ON gs.id = r.group_set_id WHERE r.id = ?`,
+      [id]
+    )
     if (!r) return null
-    const [files, { rows: inspections }, { rows: finality }, passwords] = await Promise.all([
+    const [files, { rows: inspections }, passwords] = await Promise.all([
       NzbService.files(id),
       db.rawQuery(`SELECT kind, result, created_at FROM release_inspections WHERE release_id = ? ORDER BY candidate_id`, [id]),
-      db.rawQuery(`SELECT release_final(?, ?, ?, ?, ?, ?) AS final`, [r.group_id, r.first_article, r.last_article, r.posted_at, r.last_posted_at, r.complete]),
       NzbService.passwords(id),
     ])
     return {
       id: Number(r.id),
       name: r.name,
-      source: r.source,
-      poster: r.poster,
-      groups: r.groups,
+      source: sourceName(r.source),
+      poster: r.poster ?? null,
+      groups: r.groups ?? [],
       postedAt: r.posted_at,
       fileTotal: r.file_total,
       fileCount: Number(r.files),
@@ -224,11 +250,10 @@ class SearchService {
       segments: Number(r.segments),
       readable: r.readable,
       complete: r.complete,
-      final: finality[0].final,
+      final: r.settled_at !== null,
       hasPar2: r.has_par2,
       category: r.category,
       firstSeenAt: r.first_seen_at,
-      lastSeenAt: r.last_seen_at,
       settledAt: r.settled_at,
       passworded: r.passworded,
       /** Known passwords (spots, their NZBs), the most trusted first; the first is in the NZB. */
@@ -294,9 +319,9 @@ class SearchService {
     // History by articles; the pass: the history ranges before those deferred behind them (priority
     // 2000 apart; e.g. groups added later that go first).
     const { rows: ranges } = await db.rawQuery(
-      `WITH h AS (SELECT status, priority, end_id - start_id + 1 AS n FROM ranges WHERE priority >= 0),
+      `WITH h AS (SELECT status, priority, end_id - start_id + 1 AS n FROM ranges WHERE priority >= 0 AND status <> 'ancient'),
             pass AS (SELECT (floor(coalesce(min(priority) FILTER (WHERE status <> 'done'), 0) / 2000) + 1) * 2000 AS below FROM h)
-       SELECT (SELECT count(*) FILTER (WHERE status = 'done') FROM ranges) AS done, (SELECT count(*) FROM ranges) AS total,
+       SELECT (SELECT count(*) FILTER (WHERE status = 'done') FROM ranges) AS done, (SELECT count(*) FROM ranges WHERE status <> 'ancient') AS total,
               coalesce(sum(n) FILTER (WHERE status = 'done'), 0) AS history_done, coalesce(sum(n), 0) AS history,
               coalesce(sum(n) FILTER (WHERE status = 'done' AND priority < (SELECT below FROM pass)), 0) AS pass_done,
               coalesce(sum(n) FILTER (WHERE priority < (SELECT below FROM pass)), 0) AS pass,

app/services/settler.ts

@@ -1,48 +1,35 @@
 import db from '@adonisjs/lucid/services/db'
 
-/** Releases unchanged for this long are settled (ranges are read in order; the merger adds within minutes). */
-const QUIET_HOURS = 6
-/**
- * History: a release whose last post is QUIET_HOURS old gets nothing new from the tail, only from the
- * neighbouring ranges' fragments, which the merger adds within minutes: settled once unchanged this long.
- * Waiting QUIET_HOURS after it was read held every history release back from phase 2 for six hours after
- * each reset (the hourly sweeps' resets kept the prober idle; LOG.md 2026-10-12 (13)).
- */
-const QUIET_HISTORY_MINUTES = 30
-const BATCH = 50_000
+/** Hours a complete provisional release stays open for its window's continuation before it is settled here. */
+export const SETTLE_COMPLETE_HOURS = 6
 
 /**
- * Settles releases (REBUILD.md, part 8): once nothing has been added for
- * QUIET_HOURS (QUIET_HISTORY_MINUTES once its last post is that old), its completeness is decided: complete when every file the
- * [i/N] counter announced was found, unknown without a counter (phase 2
- * decides those by their par2 sets: app/services/par2_completeness.ts).
- * Phase 2 picks its work from settled releases.
+ * Provisional releases (REBUILD.md, "Voyager 3: window indexing", 2): a release the window hadn't closed on at a
+ * lease's end is written with `settled_at` null, for the window's continuation to finish. A continuation may
+ * never come (the range-lease path of the first run left 109k such rows, LOG.md (118); a group's next lease is
+ * hours away), and the catalog shows the release as still collecting. A complete one — every file its counter
+ * announced found — can't gain a counted file any more, so once it has been quiet for SETTLE_COMPLETE_HOURS it is
+ * settled here, as the finish would: `settled_at` set (the judge's trigger fires on it), its waiting probe
+ * candidates made claimable. Incomplete ones keep waiting: the continuation adds their files.
  */
 class Settler {
-  /** Up to `batches` batches of BATCH (all there are without). */
-  async run(batches = Infinity): Promise<number> {
-    let settled = 0
-    for (let batch = 0; batch < batches; batch++) {
-      // Their probe candidates become claimable (Prober.claim), the unreadable ones first.
-      const { rows } = await db.rawQuery(
-        `WITH settled AS (
-           UPDATE releases SET settled_at = now(), complete = CASE WHEN file_total IS NULL THEN complete ELSE files >= file_total END
-           WHERE id IN (SELECT id FROM releases WHERE settled_at IS NULL AND last_seen_at < now() - make_interval(mins => ?)
-             AND (last_seen_at < now() - make_interval(hours => ?) OR coalesce(last_posted_at, posted_at) < now() - make_interval(hours => ?)) LIMIT ?)
-           RETURNING id, readable, files, bytes, poster, sidecar_only),
-         marked AS (UPDATE probe_candidates pc SET ready = true, unreadable = NOT s.readable FROM settled s WHERE pc.release_id = s.id),
-         -- A small one of sidecars only (the batch store's sidecar_only) may be an upload's .nfo or .nzb that came apart
-         -- from it: the merger's sidecar fold looks at them.
-         queued AS (INSERT INTO fold_candidates (release_id) SELECT id FROM settled WHERE sidecar_only AND files <= 3 AND bytes < 1000000 AND poster IS NOT NULL
-           ON CONFLICT DO NOTHING)
-         SELECT count(*)::int AS n FROM settled`,
-        [QUIET_HISTORY_MINUTES, QUIET_HOURS, QUIET_HOURS, BATCH]
+  async settleComplete(quietHours = SETTLE_COMPLETE_HOURS): Promise<number> {
+    return db.transaction(async (trx) => {
+      const { rows } = await trx.rawQuery(
+        `UPDATE releases SET settled_at = now()
+         WHERE settled_at IS NULL AND complete
+           AND first_seen_at < now() - make_interval(hours => ?) AND last_posted_at < now() - make_interval(hours => ?)
+         RETURNING id`,
+        [quietHours, quietHours]
       )
-      const n = rows[0].n
-      settled += n
-      if (n < BATCH) return settled
-    }
-    return settled
+      if (rows.length > 0)
+        await trx.rawQuery(
+          `UPDATE probe_candidates pc SET ready = true, unreadable = NOT r.readable FROM releases r
+           WHERE pc.release_id = ANY(?::bigint[]) AND r.id = pc.release_id AND pc.status = 'waiting' AND NOT pc.ready`,
+          [rows.map((r: { id: string }) => Number(r.id))]
+        )
+      return rows.length
+    })
   }
 }
 

app/services/sidecar_folder.ts

@@ -56,7 +56,7 @@ export class SidecarFolder {
     // A candidate whose release went (merged, deleted) leaves with its batch: a separate scan for those over the whole
     // queue took 5+ min a round at 13M (LOG.md 2026-10-17 (72)).
     const { rows: queued } = await db.rawQuery(
-      `SELECT f.release_id AS queued_id, r.id, r.release_key, r.group_id, r.poster, r.posted_at, r.files, r.file_total, r.readable
+      `SELECT f.release_id AS queued_id, r.id, r.group_id, r.poster_id, r.posted_at, r.files, r.file_total, r.readable
        FROM fold_candidates f LEFT JOIN releases r ON r.id = f.release_id ORDER BY f.created_at LIMIT ?`,
       [BATCH]
     )
@@ -70,10 +70,10 @@ export class SidecarFolder {
       round.sidecars++
       if (!c.posted_at) continue
       const { rows: siblings } = await db.rawQuery(
-        `SELECT r.id, r.release_key, r.group_id, r.files, r.file_total, r.posted_at, r.readable, r.name FROM releases r
+        `SELECT r.id, r.group_id, r.files, r.file_total, r.posted_at, r.readable, r.name FROM releases r
          WHERE r.posted_at BETWEEN ?::timestamptz - make_interval(mins => ?) AND ?::timestamptz + make_interval(mins => ?)
-           AND r.group_id = ? AND r.poster = ? AND r.id <> ? AND r.files > ?`,
-        [c.posted_at, WINDOW_MINUTES, c.posted_at, WINDOW_MINUTES, c.group_id, c.poster, c.id, c.files]
+           AND r.group_id = ? AND r.poster_id = ? AND r.id <> ? AND r.files > ?`,
+        [c.posted_at, WINDOW_MINUTES, c.posted_at, WINDOW_MINUTES, c.group_id, c.poster_id, c.id, c.files]
       )
       const sibling = pickSibling(files, c.file_total ? Number(c.file_total) : null, siblings.map((s: any) => ({ ...s, id: Number(s.id), files: Number(s.files), file_total: s.file_total === null ? null : Number(s.file_total) })))
       if (!sibling) continue

app/services/spot_inspector.ts

@@ -9,6 +9,9 @@ import { decodeSpotBinary } from '#lib/spotnet/index'
 import { parseNzb, type ParsedNzb } from '#lib/nzb/index'
 import { parseSubject } from '#lib/headers/index'
 import { releaseIdentity } from '#lib/release/index'
+import { identityKey, releaseKey } from '#lib/release/keys'
+import { SOURCES } from '#lib/release/source'
+import CatalogRefs from '#services/catalog_refs'
 
 /** Spots claimed per round, newest first. */
 const BATCH = 20
@@ -251,8 +254,8 @@ export class SpotInspector {
     }
     if (wanted.length === 0) return 0
     const { rows } = await db.rawQuery(
-      `SELECT id, identity_key, group_id, posted_at FROM releases WHERE identity_key = ANY(?) AND group_id = ANY(?)`,
-      [[...new Set(wanted.map((w) => w.key))], [...new Set(wanted.flatMap((w) => w.groups))]]
+      `SELECT id, encode(identity_key, 'hex') AS identity_key, group_id, posted_at FROM releases WHERE identity_key = ANY(?::bytea[]) AND group_id = ANY(?)`,
+      [[...new Set(wanted.map((w) => w.key))].map((k) => identityKey(k)), [...new Set(wanted.flatMap((w) => w.groups))]]
     )
     const byKey = new Map<string, { id: number; group: number; at: number | null }[]>()
     for (const r of rows) {
@@ -275,7 +278,7 @@ export class SpotInspector {
     await db.transaction(async (trx) => {
       // A release of the spot's own goes now that the header pass built its files' releases.
       if (spot.release_id) {
-        const own = await trx.from('releases').where('id', spot.release_id).where('source', 'spot').select('id').first()
+        const own = await trx.from('releases').where('id', spot.release_id).where('source', SOURCES.spot).select('id').first()
         if (own) {
           await trx.from('release_spots').where('release_id', own.id).delete()
           await trx.from('release_passwords').where('release_id', own.id).delete()
@@ -335,16 +338,19 @@ export class SpotInspector {
     const groups = [...new Set(files.flatMap((f: any) => f.groups ?? []))]
     const groupId = groups.length > 0 ? await this.groupId(groups[0]) : null
     const key = createHash('sha1').update(`spot\n${spotId}`).digest('hex')
+    // Its poster and group set by id (voyager 3, migration 6000), as the batch store stores its releases'.
+    const posterId = spot.poster ? await CatalogRefs.posterId(db, spot.poster) : null
+    const groupSetId = await CatalogRefs.groupSetId(db, groups)
     const [row] = await db
       .table('releases')
       .insert({
-        release_key: `spot:${key}`,
-        identity_key: key,
+        release_key: releaseKey(key, groupId, null),
+        identity_key: identityKey(key),
         group_id: groupId,
+        group_set_id: groupSetId,
+        poster_id: posterId,
         name: spot.title.slice(0, 1000),
-        source: 'spot',
-        poster: spot.poster,
-        groups: JSON.stringify(groups),
+        source: SOURCES.spot,
         posted_at: spot.posted_at,
         last_posted_at: spot.posted_at,
         files: files.length,
@@ -363,8 +369,8 @@ export class SpotInspector {
     const releaseId = Number(row.id)
     await db.transaction(async (trx) => {
       await trx.table('release_spots').insert({ release_id: releaseId, spot_id: spotId, files: files.length }).onConflict(['release_id', 'spot_id']).ignore()
-      // Its poster for poster search, as the batch store adds its releases' (migration 4000).
-      if (spot.poster) await trx.rawQuery(`INSERT INTO posters (poster) VALUES (left(?, 200)) ON CONFLICT DO NOTHING`, [spot.poster])
+      // Its title, searchable (release_texts).
+      await trx.rawQuery(`INSERT INTO release_texts (release_id, kind, text) VALUES (?, 1, left(?, 255)) ON CONFLICT DO NOTHING`, [releaseId, spot.title])
       if (spot.password) {
         await trx.table('release_passwords').insert({ release_id: releaseId, password: spot.password, source: spot.password_source ?? 'spot-text', spot_id: spotId }).onConflict(['release_id', 'password']).ignore()
       }

app/services/spotter.ts

@@ -1,5 +1,6 @@
 import db from '@adonisjs/lucid/services/db'
 import logger from '@adonisjs/core/services/logger'
+import { SOURCES } from '#lib/release/source'
 import env from '#start/env'
 import coordinatorConfig from '#config/coordinator'
 import { MissingArticle, XoverClient } from '#lib/nntp/index'
@@ -183,7 +184,7 @@ export class Spotter {
       await trx.from('release_spots').where('spot_id', targetId).delete()
       await trx.from('release_passwords').where('spot_id', targetId).delete()
       // A release the spot made of its own goes with it.
-      if (target.release_id) await trx.from('releases').where('id', target.release_id).where('source', 'spot').delete()
+      if (target.release_id) await trx.from('releases').where('id', target.release_id).where('source', SOURCES.spot).delete()
     })
     logger.info(`Spot "${target.title}" removed by ${personal ? 'its poster' : 'a moderator'}.`)
   }

app/services/status.ts

@@ -0,0 +1,343 @@
+import db from '@adonisjs/lucid/services/db'
+import { readdir, stat, statfs } from 'node:fs/promises'
+import { join } from 'node:path'
+import coordinatorConfig from '#config/coordinator'
+import GroupLeases from '#services/group_leases'
+import type { Lane } from '#lib/edge/index'
+
+export interface RangeBand {
+  band: string
+  planned: number
+  plannedArticles: number
+  leased: number
+  done: number
+  doneArticles: number
+  failed: number
+}
+
+export interface HeldGroup {
+  node: string
+  group: string
+  lane: Lane
+  range: { start: number; end: number; next: number } | null
+  heartbeatSeconds: number
+  snapshotNext: number | null
+  snapshotSeconds: number | null
+  snapshotMb: number | null
+  rangesDone: number
+  headers: number
+}
+
+export interface SnapshotFile {
+  group: string
+  lane: Lane
+  mb: number
+  hours: number
+}
+
+export interface Status {
+  at: string
+  catalog: {
+    tablespace: string
+    gb: number
+    capGb: number
+    full: boolean
+    freeGb: number
+  }
+  bands: RangeBand[]
+  lastHour: {
+    indexedArticles: number
+    doneRanges: number
+    storedArticles: number
+    storedRanges: number
+  }
+  corpus: { pendingRanges: number; pendingArticles: number }
+  spool: { ranges: number; chunks: number }
+  tail: { groups: number; ranges: number; articles: number }
+  /** The 'ancient' ranges: articles below the low watermark the groups were planned at (expired, never fetched). */
+  ancient: { groups: number; articles: number }
+  held: HeldGroup[]
+  snapshots: { count: number; gb: number; oldestHours: number | null; largest: SnapshotFile[] }
+  releases: { estimated: number; provisional: number }
+  /** `ready` and `waiting` are the planner's estimates (the last analyze), the others counts. */
+  probes: { ready: number; waiting: number; pendingWaiting: number; pendingNamed: number }
+  judge: { dirty: number }
+  nodes: { name: string; lastSeenSeconds: number | null }[]
+}
+
+const seconds = (since: Date | string | null) =>
+  since ? Math.round((Date.now() - new Date(since).getTime()) / 1000) : null
+
+/**
+ * What the indexer is doing, in one object (commands/coordinator_status.ts prints it): the catalog's room on the
+ * NVMe, the ranges by band (the tail, the capture windows, the history), the last hour's rates, the held groups
+ * with their snapshots, the phase-2 backlogs. Every number comes from the database, the snapshot directory or the
+ * filesystem; nothing is kept between calls.
+ */
+class StatusService {
+  async collect(): Promise<Status> {
+    const tail = coordinatorConfig.tail.priority
+    const cap = coordinatorConfig.catalogMaxNvmeGb
+    // The catalog's tablespace only (CatalogSpace measures it every minute anyway): pg_database_size walks the
+    // chunk partitions and the probe cache on the HDD, minutes on homelab.
+    const { rows: space } = await db.rawQuery(
+      `SELECT coalesce(nullif(current_setting('default_tablespace'), ''), 'pg_default') AS tablespace,
+              pg_tablespace_size(coalesce(nullif(current_setting('default_tablespace'), ''), 'pg_default')) AS bytes`
+    )
+    const tablespaceGb = Number(space[0].bytes) / 1e9
+    const fs = await statfs('/').catch(() => null)
+    const freeGb = fs ? (fs.bavail * fs.bsize) / 1e9 : 0
+
+    // The ranges by band: the tail, the capture windows (priorities 0–2; the sampling ranges in other years are
+    // capture too, at their history priorities), the history by its thousand.
+    const { rows: bands } = await db.rawQuery(
+      `SELECT CASE WHEN priority <= ? THEN 'tail'
+                   WHEN priority < 0 THEN 'raised'
+                   WHEN capture AND priority <= 2 THEN 'window ' || priority
+                   WHEN priority < 1000 THEN 'history 0-999'
+                   ELSE 'history ' || (priority / 1000) * 1000 || '+' END AS band,
+              count(*) FILTER (WHERE status = 'planned') AS planned,
+              coalesce(sum(end_id - next_id + 1) FILTER (WHERE status = 'planned'), 0) AS planned_articles,
+              count(*) FILTER (WHERE status = 'leased') AS leased,
+              count(*) FILTER (WHERE status = 'done') AS done,
+              coalesce(sum(end_id - start_id + 1) FILTER (WHERE status = 'done'), 0) AS done_articles,
+              count(*) FILTER (WHERE status = 'failed') AS failed
+       FROM ranges GROUP BY 1
+       ORDER BY min(CASE WHEN priority <= ? THEN 0 WHEN priority < 0 THEN 1 WHEN capture AND priority <= 2 THEN 2 + priority
+                        WHEN priority < 1000 THEN 10 ELSE 11 + priority / 1000 END)`,
+      [tail, tail]
+    )
+    const { rows: hour } = await db.rawQuery(
+      `SELECT coalesce(sum(end_id - start_id + 1) FILTER (WHERE done_at > now() - interval '1 hour'), 0) AS indexed,
+              count(*) FILTER (WHERE done_at > now() - interval '1 hour') AS done_ranges,
+              coalesce(sum(end_id - start_id + 1) FILTER (WHERE stored_at > now() - interval '1 hour'), 0) AS stored,
+              count(*) FILTER (WHERE stored_at > now() - interval '1 hour') AS stored_ranges,
+              count(*) FILTER (WHERE capture AND stored_at IS NULL AND status <> 'done') AS corpus_pending,
+              coalesce(sum(end_id - start_id + 1) FILTER (WHERE capture AND stored_at IS NULL AND status <> 'done'), 0) AS corpus_pending_articles,
+              count(*) FILTER (WHERE status IN ('planned', 'leased') AND spooled_chunks > 0) AS spooled_ranges,
+              coalesce(sum(spooled_chunks) FILTER (WHERE status IN ('planned', 'leased')), 0) AS spooled_chunks,
+              count(DISTINCT group_id) FILTER (WHERE status = 'planned' AND priority <= ?) AS tail_groups,
+              count(*) FILTER (WHERE status = 'planned' AND priority <= ?) AS tail_ranges,
+              coalesce(sum(end_id - next_id + 1) FILTER (WHERE status = 'planned' AND priority <= ?), 0) AS tail_articles,
+              count(DISTINCT group_id) FILTER (WHERE status = 'ancient') AS ancient_groups,
+              coalesce(sum(end_id - start_id + 1) FILTER (WHERE status = 'ancient'), 0) AS ancient_articles
+       FROM ranges`,
+      [tail, tail, tail]
+    )
+    const h = hour[0]
+
+    const ttl = coordinatorConfig.leaseTtlSeconds
+    const { rows: held } = await db.rawQuery(
+      `SELECT n.name AS node, g.name AS "group", l.lane, l.heartbeat_at, l.snapshot_next, l.snapshot_at, l.ranges_done, l.headers,
+              r.start_id, r.end_id, r.next_id, l.group_id
+       FROM group_leases l JOIN groups g ON g.id = l.group_id JOIN nodes n ON n.id = l.node_id
+       LEFT JOIN ranges r ON r.id = l.range_id
+       WHERE l.node_id IS NOT NULL AND l.heartbeat_at >= now() - make_interval(secs => ?)
+       ORDER BY l.lane, g.name`,
+      [ttl]
+    )
+    const snapshotSizes = new Map<string, number>()
+    const files: SnapshotFile[] = []
+    const names = new Map<number, string>(
+      (await db.from('groups').select('id', 'name')).map((g) => [Number(g.id), g.name as string])
+    )
+    for (const file of await readdir(GroupLeases.dir).catch(() => [] as string[])) {
+      const m = /^(\d+)(-tail)?\.gz$/.exec(file)
+      if (!m) continue
+      const s = await stat(join(GroupLeases.dir, file)).catch(() => null)
+      if (!s) continue
+      const lane: Lane = m[2] ? 'tail' : 'history'
+      snapshotSizes.set(`${m[1]}:${lane}`, s.size)
+      files.push({
+        group: names.get(Number(m[1])) ?? m[1],
+        lane,
+        mb: s.size / 1e6,
+        hours: (Date.now() - s.mtimeMs) / 3600_000,
+      })
+    }
+    files.sort((a, b) => b.mb - a.mb)
+
+    // The probe backlog from the planner's estimates of the partial indexes (counting 11 M rows took 10 s on
+    // homelab); the rest are small partial indexes.
+    const { rows: releases } = await db.rawQuery(
+      `SELECT (SELECT reltuples::bigint FROM pg_class WHERE oid = 'releases'::regclass) AS estimated,
+              (SELECT count(*) FROM releases WHERE settled_at IS NULL) AS provisional,
+              (SELECT greatest(0, reltuples)::bigint FROM pg_class WHERE oid = 'probe_candidates_ready_idx'::regclass) AS ready,
+              (SELECT greatest(0, reltuples)::bigint FROM pg_class WHERE oid = 'probe_candidates_waiting_idx'::regclass) AS waiting,
+              (SELECT count(*) FROM pending_releases WHERE status = 0) AS pending_waiting,
+              (SELECT count(*) FROM pending_releases WHERE status = 4) AS pending_named,
+              (SELECT count(*) FROM judge_dirty) AS dirty`
+    )
+    const r = releases[0]
+    const nodes = await db.from('nodes').select('name', 'last_seen_at').orderBy('name')
+
+    return {
+      at: new Date().toISOString(),
+      catalog: {
+        tablespace: space[0].tablespace,
+        gb: tablespaceGb,
+        capGb: cap,
+        full: cap > 0 && tablespaceGb > cap,
+        freeGb,
+      },
+      bands: bands.map((b: any) => ({
+        band: b.band,
+        planned: Number(b.planned),
+        plannedArticles: Number(b.planned_articles),
+        leased: Number(b.leased),
+        done: Number(b.done),
+        doneArticles: Number(b.done_articles),
+        failed: Number(b.failed),
+      })),
+      lastHour: {
+        indexedArticles: Number(h.indexed),
+        doneRanges: Number(h.done_ranges),
+        storedArticles: Number(h.stored),
+        storedRanges: Number(h.stored_ranges),
+      },
+      corpus: {
+        pendingRanges: Number(h.corpus_pending),
+        pendingArticles: Number(h.corpus_pending_articles),
+      },
+      spool: { ranges: Number(h.spooled_ranges), chunks: Number(h.spooled_chunks) },
+      ancient: { groups: Number(h.ancient_groups), articles: Number(h.ancient_articles) },
+      tail: {
+        groups: Number(h.tail_groups),
+        ranges: Number(h.tail_ranges),
+        articles: Number(h.tail_articles),
+      },
+      held: held.map((l: any) => {
+        const size = snapshotSizes.get(`${l.group_id}:${l.lane}`)
+        return {
+          node: l.node,
+          group: l.group,
+          lane: l.lane,
+          range:
+            l.start_id !== null
+              ? { start: Number(l.start_id), end: Number(l.end_id), next: Number(l.next_id) }
+              : null,
+          heartbeatSeconds: seconds(l.heartbeat_at) ?? 0,
+          snapshotNext: l.snapshot_next === null ? null : Number(l.snapshot_next),
+          snapshotSeconds: seconds(l.snapshot_at),
+          snapshotMb: size === undefined ? null : size / 1e6,
+          rangesDone: Number(l.ranges_done),
+          headers: Number(l.headers),
+        }
+      }),
+      snapshots: {
+        count: files.length,
+        gb: files.reduce((sum, f) => sum + f.mb, 0) / 1000,
+        oldestHours: files.length > 0 ? Math.max(...files.map((f) => f.hours)) : null,
+        largest: files.slice(0, 5),
+      },
+      releases: { estimated: Math.max(0, Number(r.estimated)), provisional: Number(r.provisional) },
+      probes: {
+        ready: Number(r.ready),
+        waiting: Number(r.waiting),
+        pendingWaiting: Number(r.pending_waiting),
+        pendingNamed: Number(r.pending_named),
+      },
+      judge: { dirty: Number(r.dirty) },
+      nodes: nodes.map((n) => ({ name: n.name, lastSeenSeconds: seconds(n.last_seen_at) })),
+    }
+  }
+
+  /** The status as a text panel. */
+  render(s: Status): string {
+    const n = (v: number) => Math.round(v).toLocaleString('en-US')
+    const b = (v: number) =>
+      v >= 1e9
+        ? `${(v / 1e9).toFixed(2)} B`
+        : v >= 1e6
+          ? `${(v / 1e6).toFixed(1)} M`
+          : v >= 1e3
+            ? `${(v / 1e3).toFixed(0)} k`
+            : String(v)
+    const age = (sec: number | null) =>
+      sec === null
+        ? '-'
+        : sec < 90
+          ? `${sec} s`
+          : sec < 5400
+            ? `${Math.round(sec / 60)} min`
+            : `${(sec / 3600).toFixed(1)} h`
+    const lines: string[] = []
+    lines.push(`Status at ${s.at}`)
+    lines.push('')
+    lines.push(
+      `Catalog: ${s.catalog.gb.toFixed(1)} GB in tablespace ${s.catalog.tablespace}${s.catalog.capGb ? ` of ${s.catalog.capGb} GB${s.catalog.full ? ' (FULL: history paused)' : ''}` : ' (no cap)'}; ${s.catalog.freeGb.toFixed(1)} GB free on /`
+    )
+    lines.push(
+      `Last hour: ${b(s.lastHour.indexedArticles)} articles indexed in ${n(s.lastHour.doneRanges)} ranges (${n(s.lastHour.indexedArticles / 3600)}/s); corpus fill ${b(s.lastHour.storedArticles)} in ${n(s.lastHour.storedRanges)} ranges, ${n(s.corpus.pendingRanges)} capture ranges (${b(s.corpus.pendingArticles)}) still to store`
+    )
+    lines.push(
+      `Spool: ${n(s.spool.chunks)} chunks in ${n(s.spool.ranges)} ranges. Tail: ${b(s.tail.articles)} planned in ${n(s.tail.ranges)} ranges of ${n(s.tail.groups)} groups. Ancient: ${b(s.ancient.articles)} articles below the low watermark of ${n(s.ancient.groups)} groups`
+    )
+    lines.push('')
+    const table = (header: string[], rows: string[][]) => {
+      const widths = header.map((h, i) => Math.max(h.length, ...rows.map((r) => r[i].length)))
+      const line = (r: string[]) =>
+        r.map((c, i) => (i === 0 ? c.padEnd(widths[i]) : c.padStart(widths[i]))).join('  ')
+      lines.push(line(header))
+      for (const r of rows) lines.push(line(r))
+    }
+    lines.push('Ranges by band:')
+    table(
+      ['band', 'planned', 'articles', 'leased', 'done', 'articles', 'failed'],
+      s.bands.map((x) => [
+        x.band,
+        n(x.planned),
+        b(x.plannedArticles),
+        n(x.leased),
+        n(x.done),
+        b(x.doneArticles),
+        n(x.failed),
+      ])
+    )
+    lines.push('')
+    lines.push(`Held groups (${s.held.length}):`)
+    if (s.held.length > 0)
+      table(
+        [
+          'group',
+          'lane',
+          'node',
+          'range',
+          'at',
+          'heartbeat',
+          'snapshot',
+          'size',
+          'done',
+          'headers',
+        ],
+        s.held.map((x) => [
+          x.group,
+          x.lane,
+          x.node,
+          x.range ? `${x.range.start}-${x.range.end}` : '-',
+          x.range ? String(x.range.next) : '-',
+          age(x.heartbeatSeconds),
+          x.snapshotNext === null ? '-' : `${x.snapshotNext} (${age(x.snapshotSeconds)})`,
+          x.snapshotMb === null ? '-' : `${x.snapshotMb.toFixed(0)} MB`,
+          n(x.rangesDone),
+          b(x.headers),
+        ])
+      )
+    lines.push('')
+    lines.push(
+      `Snapshots: ${n(s.snapshots.count)} files, ${s.snapshots.gb.toFixed(2)} GB${s.snapshots.oldestHours === null ? '' : `, oldest ${s.snapshots.oldestHours.toFixed(1)} h`}${s.snapshots.largest.length > 0 ? `; largest ${s.snapshots.largest.map((f) => `${f.group}${f.lane === 'tail' ? ' (tail)' : ''} ${f.mb.toFixed(0)} MB`).join(', ')}` : ''}`
+    )
+    lines.push(
+      `Releases: ~${n(s.releases.estimated)}, ${n(s.releases.provisional)} provisional. Judge: ${n(s.judge.dirty)} dirty`
+    )
+    lines.push(
+      `Probes: ~${n(s.probes.ready)} ready of ~${n(s.probes.waiting)} waiting; pending ladder ${n(s.probes.pendingWaiting)} waiting, ${n(s.probes.pendingNamed)} named`
+    )
+    lines.push(
+      `Nodes: ${s.nodes.map((x) => `${x.name} (${age(x.lastSeenSeconds)})`).join(', ') || 'none'}`
+    )
+    return lines.join('\n')
+  }
+}
+
+export default new StatusService()

commands/coordinator_run.ts

@@ -3,48 +3,46 @@ import type { CommandOptions } from '@adonisjs/core/types/ace'
 import { setTimeout as sleep } from 'node:timers/promises'
 import { Prefetcher } from '#services/prefetcher'
 import Planner from '#services/planner'
-import FragmentMerger from '#services/fragment_merger'
-import Settler from '#services/settler'
-import RangeBlocks from '#services/range_blocks'
 import Leases from '#services/leases'
+import Settler from '#services/settler'
 import NntpPool from '#services/nntp_pool'
 import { NntpShare } from '#services/nntp_shares'
 import coordinatorConfig from '#config/coordinator'
 
 const TICK_MS = 2000
 const PLAN_EVERY_MS = 60 * 60 * 1000
-const MERGE_EVERY_MS = 60 * 1000
-const MERGE_BUDGET_MS = 30 * 1000
 const STOP_WAIT_MS = 15 * 1000
-/**
- * Settling in rounds of at most SETTLE_BATCHES batches every few minutes: once an hour, ~3M releases rewrote their
- * rows and index entries in one go (2.65M in 30 min), the NVMe saturated and the batch stores queued on their locks
- * (LOG.md 2026-10-16 (67)).
- */
-const SETTLE_EVERY_MS = 5 * 60 * 1000
-const SETTLE_BATCHES = 4
-/** range_blocks (release_final) after ranges finish. */
-const BLOCKS_EVERY_MS = 5 * 60 * 1000
+/** Leases whose worker went quiet after reading everything are finished every few minutes. */
+const ABANDONED_EVERY_MS = 5 * 60 * 1000
 
 /**
- * The coordinator's background work (REBUILD.md, part 1): the prefetcher,
- * merging fragments of finished ranges every minute, and every hour
- * planning ranges for new articles and settling quiet releases. The API (leases, stream,
- * batches) runs in the web server.
+ * The coordinator's background work (REBUILD.md, part 1): the prefetcher, the tail's ranges every few minutes,
+ * new groups' history every hour, and abandoned leases. The API (leases, stream, batches) runs in the web server.
+ * Voyager 3: the settler and the fragment merger are gone (the window finishes a release once; the block merger
+ * joins edge releases).
  */
 export default class CoordinatorRun extends BaseCommand {
   static commandName = 'coordinator:run'
   static description = 'Prefetches ranges into the spool and plans new ones'
   static options: CommandOptions = { startApp: true, staysAlive: true }
 
-  @flags.string({ description: 'This process\'s share of the prefetching, "i/n" (ranges with id % n = i - 1); only 1/n plans, merges and settles' })
+  @flags.string({
+    description:
+      'This process\'s share of the prefetching, "i/n" (ranges with id % n = i - 1); only 1/n plans, merges and settles',
+  })
   declare shard?: string
 
   async run() {
     const [index, count] = (this.shard ?? '1/1').split('/').map(Number)
-    if (!(count >= 1 && index >= 1 && index <= count)) throw new Error(`--shard: "i/n" with 1 <= i <= n, not "${this.shard}"`)
+    if (!(count >= 1 && index >= 1 && index <= count))
+      throw new Error(`--shard: "i/n" with 1 <= i <= n, not "${this.shard}"`)
     const prefetcher = new Prefetcher({ index: index - 1, count })
-    const share = new NntpShare(count > 1 ? `coordinator ${index}/${count}` : 'coordinator', coordinatorConfig.nntpShares.coordinatorPriority, coordinatorConfig.nntpMinConnections, coordinatorConfig.nntpConnections)
+    const share = new NntpShare(
+      count > 1 ? `coordinator ${index}/${count}` : 'coordinator',
+      coordinatorConfig.nntpShares.coordinatorPriority,
+      coordinatorConfig.nntpMinConnections,
+      coordinatorConfig.nntpConnections
+    )
     const main = index === 1
     let stopping = false
     let stopped: () => void
@@ -56,7 +54,11 @@ export default class CoordinatorRun extends BaseCommand {
       prefetcher.stop()
       await done
     })
-    this.logger.info(count > 1 ? `Coordinator started (shard ${index}/${count}${main ? '' : ', prefetching only'}).` : 'Coordinator started.')
+    this.logger.info(
+      count > 1
+        ? `Coordinator started (shard ${index}/${count}${main ? '' : ', prefetching only'}).`
+        : 'Coordinator started.'
+    )
     // The pool's size: its share of the account's connections, the prober the rest (app/services/nntp_shares.ts).
     // Before the first connection: until then it's NNTP_CONNECTIONS.
     const first = await NntpPool.share(share)
@@ -69,11 +71,16 @@ export default class CoordinatorRun extends BaseCommand {
         } catch (error: any) {
           this.logger.error(`NNTP share failed: ${error.message}`)
         }
-        for (let waited = 0; waited < coordinatorConfig.nntpShares.everyMs && !stopping; waited += TICK_MS) await sleep(TICK_MS)
+        for (
+          let waited = 0;
+          waited < coordinatorConfig.nntpShares.everyMs && !stopping;
+          waited += TICK_MS
+        )
+          await sleep(TICK_MS)
       }
     })()
-    // Two loops: the prefetcher's ticks don't wait for merging or planning (a merge round
-    // of up to MERGE_BUDGET_MS left the connections idle once their targets were fetched).
+    // The prefetcher's ticks in a loop of their own: planning doesn't hold them (a merge round of up to 30 s once
+    // left the connections idle once their targets were fetched).
     const prefetching = (async () => {
       while (!stopping) {
         try {
@@ -84,24 +91,31 @@ export default class CoordinatorRun extends BaseCommand {
         await sleep(TICK_MS)
       }
     })()
-    // range_blocks in a loop of its own: a settling or merging round doesn't hold release_final back.
-    const blocks = (async () => {
+    const abandoned = (async () => {
       while (main && !stopping) {
         try {
-          const abandoned = await Leases.finishAbandoned()
-          if (abandoned > 0) this.logger.info(`Finished ${abandoned} ranges read to their end whose worker went quiet.`)
-          await RangeBlocks.rebuild()
+          const finished = await Leases.finishAbandoned()
+          if (finished > 0)
+            this.logger.info(
+              `Finished ${finished} ranges read to their end whose worker went quiet.`
+            )
         } catch (error: any) {
-          this.logger.error(`Range blocks failed: ${error.message}`)
+          this.logger.error(`Finishing abandoned leases failed: ${error.message}`)
         }
-        for (let waited = 0; waited < BLOCKS_EVERY_MS && !stopping; waited += TICK_MS) await sleep(TICK_MS)
+        try {
+          const settled = await Settler.settleComplete()
+          if (settled > 0)
+            this.logger.info(`Settled ${settled} complete provisional releases quiet for hours.`)
+        } catch (error: any) {
+          this.logger.error(`Settling complete releases failed: ${error.message}`)
+        }
+        for (let waited = 0; waited < ABANDONED_EVERY_MS && !stopping; waited += TICK_MS)
+          await sleep(TICK_MS)
       }
     })()
     // History for new groups hourly; the tail every few minutes.
     let plannedAt = 0
     let tailedAt = 0
-    let mergedAt = 0
-    let settledAt = 0
     while (!stopping) {
       try {
         if (main && Date.now() - plannedAt >= PLAN_EVERY_MS) {
@@ -109,45 +123,25 @@ export default class CoordinatorRun extends BaseCommand {
           const { groups, ranges } = await Planner.planAll()
           if (ranges > 0) this.logger.info(`Planned ${ranges} ranges over ${groups} groups.`)
         }
-        if (main && Date.now() - settledAt >= SETTLE_EVERY_MS) {
-          settledAt = Date.now()
-          const settled = await Settler.run(SETTLE_BATCHES)
-          if (settled > 0) this.logger.info(`Settled ${settled} releases.`)
-        }
         if (main && Date.now() - tailedAt >= coordinatorConfig.tail.everySeconds * 1000) {
           tailedAt = Date.now()
-          const { groups, ranges } = await Planner.planTail()
-          if (ranges > 0) this.logger.info(`Tail: ${ranges} ranges for ${groups} groups.`)
-        }
-        if (main && Date.now() - mergedAt >= MERGE_EVERY_MS) {
-          mergedAt = Date.now()
-          // Rounds until a whole pass over the file keys decides nothing more, or MERGE_BUDGET_MS.
-          // (Going on while only waiting files were left kept Postgres busy all the time.)
-          const totals = { completed: 0, incomplete: 0, waiting: 0 }
-          let decidedThisPass = 0
-          while (!stopping && Date.now() - mergedAt < MERGE_BUDGET_MS) {
-            const merged = await FragmentMerger.run()
-            totals.completed += merged.completed
-            totals.incomplete += merged.incomplete
-            totals.waiting = merged.waiting
-            decidedThisPass += merged.completed + merged.incomplete
-            if (merged.wrapped) {
-              if (decidedThisPass === 0) break
-              decidedThisPass = 0
-            }
-          }
-          if (totals.completed || totals.incomplete) this.logger.info(`Fragments: ${totals.completed} files completed, ${totals.incomplete} incomplete, ${totals.waiting} waiting.`)
+          const { groups, ranges, extended } = await Planner.planTail()
+          if (ranges + extended > 0)
+            this.logger.info(
+              `Tail: ${ranges} ranges added, ${extended} grown, for ${groups} groups.`
+            )
         }
       } catch (error: any) {
         this.logger.error(`Coordinator round failed: ${error.message}`)
       }
       await sleep(TICK_MS)
     }
-    await Promise.all([prefetching, blocks, sharing])
+    await Promise.all([prefetching, abandoned, sharing])
     // The chunks in flight, for at most STOP_WAIT_MS (a connection the server holds back can take minutes); then closed.
     const waitUntil = Date.now() + STOP_WAIT_MS
     while (prefetcher.active > 0 && Date.now() < waitUntil) await sleep(200)
-    if (prefetcher.active > 0) this.logger.info(`Closing ${prefetcher.active} connections still fetching.`)
+    if (prefetcher.active > 0)
+      this.logger.info(`Closing ${prefetcher.active} connections still fetching.`)
     await NntpPool.close()
     await share.leave().catch(() => {})
     this.logger.info('Coordinator stopped.')

commands/coordinator_status.ts

@@ -0,0 +1,22 @@
+import { BaseCommand, flags } from '@adonisjs/core/ace'
+import type { CommandOptions } from '@adonisjs/core/types/ace'
+import StatusService from '#services/status'
+
+/**
+ * What the indexer is doing (app/services/status.ts): the catalog's room, the ranges by band, the last hour's
+ * rates, the held groups and their snapshots, the phase-2 backlogs. Reads only.
+ */
+export default class CoordinatorStatus extends BaseCommand {
+  static commandName = 'coordinator:status'
+  static description =
+    'Prints what the indexer is doing: catalog room, ranges by band, rates, held groups, backlogs'
+  static options: CommandOptions = { startApp: true }
+
+  @flags.boolean({ description: 'JSON instead of the text panel' })
+  declare json: boolean
+
+  async run() {
+    const status = await StatusService.collect()
+    this.logger.log(this.json ? JSON.stringify(status, null, 2) : StatusService.render(status))
+  }
+}

commands/edge_bench.ts

@@ -108,7 +108,7 @@ export default class EdgeBench extends BaseCommand {
     let text = 0
     let peakRss = 0
     let peakDetail = ''
-    const out = { headers: 0, releases: 0, files: 0, segments: 0, payload: 0, candidates: 0, fragments: 0, samples: 0 }
+    const out = { headers: 0, releases: 0, files: 0, segments: 0, payload: 0, candidates: 0, samples: 0 }
     const counts = new Map<string, number>()
     const keys = new Set<string>()
     let batches = 0
@@ -164,7 +164,6 @@ export default class EdgeBench extends BaseCommand {
         out.payload += r.payload.length
       }
       out.candidates += batch.candidates.length
-      out.fragments += batch.fragments.length
       out.samples += batch.samples.length
       for (const c of batch.counts) counts.set(`${c.rule} ${c.key}`.trim(), (counts.get(`${c.rule} ${c.key}`.trim()) ?? 0) + c.articles)
       const m = process.memoryUsage()
@@ -182,7 +181,7 @@ export default class EdgeBench extends BaseCommand {
     this.logger.info(`${out.headers.toLocaleString('en')} headers in ${(total / 1000).toFixed(1)} s (${Math.round(out.headers / (total / 1000)).toLocaleString('en')}/s on one core; CPU ${cpuS.toFixed(1)} s, wall ${((Date.now() - wall) / 1000).toFixed(1)} s)`)
     this.logger.info(`  inflate ${(t.inflate / 1000).toFixed(1)} s, parse ${(t.parse / 1000).toFixed(1)} s, index ${(t.index / 1000).toFixed(1)} s; peak RSS ${(peakRss / 1e6).toFixed(0)} MB (${peakDetail})`)
     this.logger.info(`  ${batches} batches of ${this.batchChunks} chunks: ${out.releases} release chunks for ${keys.size} releases (${(out.releases / Math.max(1, keys.size)).toFixed(2)} chunks per release)`)
-    this.logger.info(`  kept: ${out.releases} release chunks, ${out.files} files, ${out.segments} segments, payload ${(out.payload / 1e6).toFixed(1)} MB (${(out.payload / Math.max(1, out.segments)).toFixed(1)} B/segment); ${out.candidates} candidates, ${out.fragments} fragments, ${out.samples} samples`)
+    this.logger.info(`  kept: ${out.releases} release chunks, ${out.files} files, ${out.segments} segments, payload ${(out.payload / 1e6).toFixed(1)} MB (${(out.payload / Math.max(1, out.segments)).toFixed(1)} B/segment); ${out.candidates} candidates, ${out.samples} samples`)
     for (const [k, v] of [...counts].sort((a, b) => b[1] - a[1]).slice(0, 8)) this.logger.info(`  dropped ${k}: ${v.toLocaleString('en')} articles`)
     if (this.store) this.logger.info(`  stored ${batches} batches in ${(storeMs / 1000).toFixed(1)} s (${(storeMs / Math.max(1, out.releases)).toFixed(2)} ms per release chunk)`)
   }

commands/nntp_probe.ts

@@ -30,6 +30,14 @@ export default class NntpProbe extends BaseCommand {
   @flags.boolean({ description: 'Only the span itself and the wider ones, no bisection' })
   declare noBisect: boolean
 
+  @flags.boolean({ description: 'Without XFEATURE COMPRESS GZIP' })
+  declare plain: boolean
+
+  @flags.string({
+    description: 'Only spans from the article of these sizes, e.g. "8,16,24,32,48,64"',
+  })
+  declare sizes?: string
+
   private timeoutMs = 20_000
   private info = { count: 0, low: 0, high: 0 }
 
@@ -41,10 +49,14 @@ export default class NntpProbe extends BaseCommand {
     this.info = await client.group(this.group)
     client.close()
     this.logger.info(
-      `${this.group}: ${this.info.count} articles, ${this.info.low}–${this.info.high}.`
+      `${this.group}: ${this.info.count} articles, ${this.info.low}–${this.info.high}${this.plain ? '; plain' : '; compressed'}.`
     )
     const span = (from: number, to: number) =>
       [Math.max(this.info.low, from), Math.min(this.info.high, to)] as const
+    if (this.sizes) {
+      for (const n of this.sizes.split(',').map(Number)) await this.probe(...span(a, a + n - 1))
+      return
+    }
     this.logger.info('The span and wider ones:')
     await this.probe(...span(a, a + size - 1))
     await this.probe(...span(a, a + size * 32 - 1))
@@ -83,7 +95,7 @@ export default class NntpProbe extends BaseCommand {
     } catch (error: any) {
       const timedOut = /timed out/i.test(error.message)
       this.logger.warning(
-        `  XOVER ${from}-${to}: ${error.message} after ${Date.now() - startedAt} ms${client ? `, ${client.buffered} bytes buffered` : ''}`
+        `  XOVER ${from}-${to}: ${error.message} after ${Date.now() - startedAt} ms${client ? `, ${client.buffered} bytes buffered: ${JSON.stringify(client.bufferedHead().toString('latin1'))} (${client.bufferedHead().toString('hex')})` : ''}`
       )
       return timedOut
     } finally {
@@ -98,7 +110,7 @@ export default class NntpProbe extends BaseCommand {
       user: env.get('NNTP_USER'),
       password: env.get('NNTP_PASSWORD'),
       secure: env.get('NNTP_SECURE', true),
-      compress: true,
+      compress: !this.plain,
       timeoutMs: this.timeoutMs,
     })
   }

commands/pending_classes.ts

@@ -0,0 +1,105 @@
+import { BaseCommand, flags } from '@adonisjs/core/ace'
+import type { CommandOptions } from '@adonisjs/core/types/ace'
+import db from '@adonisjs/lucid/services/db'
+import coordinatorConfig from '#config/coordinator'
+import { describeClass } from '#lib/release/index'
+
+/**
+ * The pending ladder's classes (LOG.md (130)) on the rows from before them: `klass` (the SQL pending_class) and
+ * `target_bytes` (the first target's bytes, when its candidates carry them). The finished rows (named, exhausted,
+ * naming) in one scan, seeding each class's record, which gets its verdict (dropped) as the ladder gives it; the
+ * waiting rows in windows of created_at by their index (a scan per window starved the prober on the HDD); then
+ * pending_classes' `waiting` counted afresh from the claim's index. --report shows the record.
+ */
+export default class PendingClasses extends BaseCommand {
+  static commandName = 'pending:classes'
+  static description =
+    'Classes the pending releases from before the classes and recounts pending_classes; --report shows the record'
+  static options: CommandOptions = { startApp: true }
+
+  @flags.number({ description: 'Hours of created_at per window' })
+  declare hours: number
+
+  @flags.boolean({ description: 'Show the classes with the best and worst records, write nothing' })
+  declare report: boolean
+
+  async run() {
+    if (this.report) return this.showReport()
+    const hours = this.hours ?? 6
+    const started = Date.now()
+    // The finished rows (named, exhausted, naming; 6.5M on 2026-11-02) in one pass: a scan of the table, no index
+    // reaches them by status; they seed each class's record.
+    const { rows: seeded } = await db.rawQuery(
+      `WITH classed AS (
+         UPDATE pending_releases SET klass = pending_class(name, bytes, partial),
+           target_bytes = (SELECT min((c->>'bytes')::bigint) FROM jsonb_array_elements(candidates) c WHERE c->>'bytes' IS NOT NULL)
+         WHERE klass = 0 AND status IN (1, 3, 4)
+         RETURNING tool, group_id, klass, status),
+       upserted AS (
+         INSERT INTO pending_classes (tool, group_id, klass, probed, named)
+         SELECT tool, group_id, klass, count(*), count(*) FILTER (WHERE status IN (1, 4)) FROM classed GROUP BY 1, 2, 3
+         ON CONFLICT (tool, group_id, klass) DO UPDATE SET probed = pending_classes.probed + excluded.probed, named = pending_classes.named + excluded.named
+         RETURNING 1)
+       SELECT (SELECT count(*) FROM classed) AS classed, (SELECT count(*) FROM upserted) AS classes`
+    )
+    this.logger.info(
+      `Finished rows: ${seeded[0].classed} classed into ${seeded[0].classes} classes (${Math.round((Date.now() - started) / 1000)} s).`
+    )
+    const { pendingMinProbes, pendingMaxNamed } = coordinatorConfig.probe
+    await db.rawQuery(
+      `UPDATE pending_classes SET dropped = true, swept_at = NULL WHERE NOT dropped AND probed >= ?::int AND named::float <= probed * ?::float`,
+      [pendingMinProbes, pendingMaxNamed]
+    )
+    // The waiting rows by windows of created_at (pending_releases_waiting_idx reaches them).
+    const { rows: span } = await db.rawQuery(
+      `SELECT (SELECT min(created_at) FROM pending_releases WHERE status = 0) AS lo, (SELECT max(created_at) FROM pending_releases WHERE status = 0) AS hi`
+    )
+    if (span[0].lo) {
+      let from = new Date(span[0].lo)
+      const hi = new Date(span[0].hi)
+      let classed = 0
+      while (from <= hi) {
+        const to = new Date(from.getTime() + hours * 3600_000)
+        const result = await db.rawQuery(
+          `UPDATE pending_releases SET klass = pending_class(name, bytes, partial),
+             target_bytes = (SELECT min((c->>'bytes')::bigint) FROM jsonb_array_elements(candidates) c WHERE c->>'bytes' IS NOT NULL)
+           WHERE status = 0 AND created_at >= ? AND created_at < ? AND klass = 0`,
+          [from.toISOString(), to.toISOString()]
+        )
+        classed += result.rowCount ?? 0
+        from = to
+        this.logger.info(
+          `Waiting rows up to ${to.toISOString().slice(0, 16)}: ${classed} classed (${Math.round((Date.now() - started) / 1000)} s).`
+        )
+      }
+    }
+    // The waiting counts from the rows themselves (the claim, the discard and the batch store keep them from here).
+    await db.rawQuery(`
+      INSERT INTO pending_classes (tool, group_id, klass, waiting)
+      SELECT tool, group_id, klass, count(*) FROM pending_releases WHERE status = 0 AND finished GROUP BY 1, 2, 3
+      ON CONFLICT (tool, group_id, klass) DO UPDATE SET waiting = excluded.waiting`)
+    await db.rawQuery(
+      `UPDATE pending_classes c SET waiting = 0 WHERE waiting > 0 AND NOT EXISTS (
+         SELECT 1 FROM pending_releases p WHERE p.status = 0 AND p.finished AND p.tool = c.tool AND p.group_id = c.group_id AND p.klass = c.klass)`
+    )
+    const { rows: totals } = await db.rawQuery(
+      `SELECT count(*) AS classes, coalesce(sum(waiting), 0) AS waiting, count(*) FILTER (WHERE dropped) AS dropped FROM pending_classes`
+    )
+    this.logger.info(
+      `Done: ${totals[0].classes} classes, ${totals[0].waiting} waiting, ${totals[0].dropped} dropped.`
+    )
+  }
+
+  private async showReport() {
+    const { rows } = await db.rawQuery(
+      `SELECT c.tool, g.name AS grp, c.klass, c.probed, c.named, c.waiting, c.dropped FROM pending_classes c JOIN groups g ON g.id = c.group_id
+       WHERE c.probed >= 20 ORDER BY c.named::float / c.probed DESC, c.probed DESC`
+    )
+    const line = (r: any) =>
+      `  ${r.tool.padEnd(10)} ${r.grp.padEnd(32)} ${describeClass(Number(r.klass)).padEnd(26)} probed ${String(r.probed).padStart(7)} named ${String(Math.round((100 * r.named) / r.probed)).padStart(3)}% waiting ${String(r.waiting).padStart(8)}${r.dropped ? ' dropped' : ''}`
+    this.logger.info(`${rows.length} classes with 20+ climbs; the best:`)
+    for (const r of rows.slice(0, 25)) this.logger.info(line(r))
+    this.logger.info('The worst:')
+    for (const r of rows.slice(-25)) this.logger.info(line(r))
+  }
+}

commands/ranges_dates.ts

@@ -1,8 +1,9 @@
 import { BaseCommand, flags } from '@adonisjs/core/ace'
 import type { CommandOptions } from '@adonisjs/core/types/ace'
 import db from '@adonisjs/lucid/services/db'
+import env from '#start/env'
 import coordinatorConfig from '#config/coordinator'
-import NntpPool from '#services/nntp_pool'
+import { XoverClient } from '#lib/nntp/index'
 import { NntpShare } from '#services/nntp_shares'
 import { estimateDate, sampleGroup, type DateSample } from '#services/group_dates'
 
@@ -39,13 +40,9 @@ export default class RangesDates extends BaseCommand {
   declare estimateOnly: boolean
 
   async run() {
-    try {
-      if (!this.estimateOnly) await this.sample()
-      await this.estimate()
-      await this.report()
-    } finally {
-      await NntpPool.close()
-    }
+    if (!this.estimateOnly) await this.sample()
+    await this.estimate()
+    await this.report()
   }
 
   private groups(missing = this.missing) {
@@ -62,20 +59,26 @@ export default class RangesDates extends BaseCommand {
     const every = this.every ?? 500_000_000
     const connections = this.connections ?? 6
     const groups = await this.groups()
+    // Connections of its own, without compression: Eweka cuts a compressed reply at the first ".\r\n" in its
+    // output, which 16 of 14k small XOVERs hit (TruncatedReply; LOG.md 2026-10-25 (105)); plain, every one
+    // answers, and 32 lines are a few kB. Taken from the account as a share (the coordinator and the prober make
+    // room for them), as many as asked for.
+    const workers = Math.max(1, Math.min(connections, groups.length))
     const share = new NntpShare(
       'dates',
       coordinatorConfig.nntpShares.coordinatorPriority + 1,
-      1,
-      connections
-    )
-    const first = await NntpPool.share(share)
-    const sharing = setInterval(
-      () =>
-        NntpPool.share(share).catch((error) => this.logger.warning(`NNTP share: ${error.message}`)),
-      coordinatorConfig.nntpShares.everyMs
+      workers,
+      workers
     )
+    let open = 0
+    const sync = () =>
+      share
+        .sync(workers, open)
+        .catch((error) => this.logger.warning(`NNTP share: ${error.message}`))
+    await sync()
+    const sharing = setInterval(sync, coordinatorConfig.nntpShares.everyMs)
     this.logger.info(
-      `Sampling ${groups.length} groups over ${first.granted} connections, a sample every ${every / 1e6}M articles.`
+      `Sampling ${groups.length} groups over ${workers} plain connections, a sample every ${every / 1e6}M articles.`
     )
     const startedAt = Date.now()
     let done = 0
@@ -83,12 +86,32 @@ export default class RangesDates extends BaseCommand {
     let samples = 0
     const queue = [...groups]
     const worker = async () => {
+      let client: XoverClient | null = null
+      const use = async <T>(work: (c: XoverClient) => Promise<T>): Promise<T> => {
+        if (!client) {
+          client = await XoverClient.connect({
+            host: env.get('NNTP_HOST')!,
+            port: env.get('NNTP_PORT', 563),
+            user: env.get('NNTP_USER'),
+            password: env.get('NNTP_PASSWORD'),
+            secure: env.get('NNTP_SECURE', true),
+            compress: false,
+            timeoutMs: coordinatorConfig.nntpTimeoutMs,
+          })
+          open++
+        }
+        try {
+          return await work(client)
+        } catch (error) {
+          client.close()
+          client = null
+          open--
+          throw error
+        }
+      }
       for (let g = queue.shift(); g; g = queue.shift()) {
         try {
-          const result = await sampleGroup((work) => NntpPool.use(work), g.name, {
-            every,
-            size: 32,
-          })
+          const result = await sampleGroup(use, g.name, { every, size: 32 })
           for (const f of result.failed) this.logger.warning(`${g.name} at ${f}`)
           await db.transaction(async (trx) => {
             await trx.from('group_dates').where('group_id', g.id).delete()
@@ -116,13 +139,16 @@ export default class RangesDates extends BaseCommand {
             `${done} of ${groups.length} groups: ${samples} samples, ${commands} commands, ${((Date.now() - startedAt) / 1000).toFixed(0)} s.`
           )
       }
+      if (client) {
+        ;(client as XoverClient).close()
+        open--
+      }
     }
     try {
-      await Promise.all(
-        Array.from({ length: Math.max(1, Math.min(connections, groups.length)) }, worker)
-      )
+      await Promise.all(Array.from({ length: workers }, worker))
     } finally {
       clearInterval(sharing)
+      await share.sync(0, 0).catch(() => {})
     }
     this.logger.info(
       `Sampled ${groups.length} groups: ${samples} samples, ${commands} commands, ${((Date.now() - startedAt) / 1000).toFixed(0)} s.`

commands/ranges_merge_tail.ts

@@ -0,0 +1,23 @@
+import { BaseCommand, flags } from '@adonisjs/core/ace'
+import type { CommandOptions } from '@adonisjs/core/types/ace'
+import Planner from '#services/planner'
+
+/**
+ * Merges adjacent planned tail ranges (Planner.mergeTail): the tail planner added a sliver a group every round
+ * while nothing leased the tail; now it grows the group's open range instead, and this packs what is there.
+ */
+export default class RangesMergeTail extends BaseCommand {
+  static commandName = 'ranges:merge-tail'
+  static description = 'Merges adjacent planned tail ranges into ranges of RANGE_ARTICLES'
+  static options: CommandOptions = { startApp: true }
+
+  @flags.boolean({ description: 'Only count what would be merged' })
+  declare dryRun: boolean
+
+  async run() {
+    const { candidates, merged, dropped } = await Planner.mergeTail(this.dryRun)
+    this.logger.info(
+      `${candidates} planned tail ranges: ${dropped} ${this.dryRun ? 'would merge' : 'merged'} into ${merged} grown ranges (${candidates - dropped} left).`
+    )
+  }
+}

commands/ranges_plan.ts

@@ -14,8 +14,13 @@ export default class RangesPlan extends BaseCommand {
 
   async run() {
     try {
-      const { groups, ranges } = this.tail ? await Planner.planTail() : await Planner.planAll()
-      this.logger.info(`${ranges} ${this.tail ? 'tail ' : ''}ranges planned over ${groups} groups.`)
+      if (this.tail) {
+        const { groups, ranges, extended } = await Planner.planTail()
+        this.logger.info(`Tail: ${ranges} ranges added, ${extended} grown, over ${groups} groups.`)
+      } else {
+        const { groups, ranges } = await Planner.planAll()
+        this.logger.info(`${ranges} ranges planned over ${groups} groups.`)
+      }
     } finally {
       await NntpPool.close()
     }

commands/releases_readable.ts

@@ -0,0 +1,90 @@
+import { BaseCommand, flags } from '@adonisjs/core/ace'
+import type { CommandOptions } from '@adonisjs/core/types/ace'
+import db from '@adonisjs/lucid/services/db'
+import { isUnreadableName } from '#lib/release/index'
+
+/** Its place in worker_cursors (the last release id done), for a run that was stopped. */
+const CURSOR = 'readable'
+
+/**
+ * The readable flag of the catalog's releases by today's classifier (isUnreadableName, LOG.md (129)): rows flagged
+ * readable whose name no one can read ("Zx5AjNaK-mh7CFg6Hf8uu", "Description - Lt5nHjlo3RiuBb3C") get readable =
+ * false, so the readable filter hides them (an inner name from the prober still shows them), and their probe
+ * candidates marked unreadable, the prober's first tiers. By the release name,
+ * which is the shared file stem for most rows; not by the chunks' file names, which the edge judges from now on.
+ * Batches by id, a dry run by default; --apply writes, and goes on from its cursor when run again.
+ */
+export default class ReleasesReadable extends BaseCommand {
+  static commandName = 'releases:readable'
+  static description =
+    'Clears the readable flag of catalog rows whose name is unreadable by the current classifier; counts only without --apply'
+  static options: CommandOptions = { startApp: true }
+
+  @flags.number({ description: 'Release ids per batch' })
+  declare batch: number
+
+  @flags.number({ description: 'Stop after this id (a sample)' })
+  declare toId: number
+
+  @flags.boolean({ description: 'Write the flag; without it, count and show a sample' })
+  declare apply: boolean
+
+  @flags.boolean({ description: 'Start from the first id, not the cursor' })
+  declare restart: boolean
+
+  async run() {
+    const batch = this.batch ?? 20_000
+    const { rows: ends } = await db.rawQuery(`SELECT min(id) AS lo, max(id) AS hi FROM releases`)
+    const hi = Math.min(Number(ends[0].hi), this.toId ?? Infinity)
+    const { rows: cursors } = await db.rawQuery(`SELECT key FROM worker_cursors WHERE name = ?`, [
+      CURSOR,
+    ])
+    let after =
+      this.restart || !this.apply || !cursors[0] ? Number(ends[0].lo) - 1 : Number(cursors[0].key)
+    let looked = 0
+    let changed = 0
+    const sample: string[] = []
+    const started = Date.now()
+    while (after < hi) {
+      const to = Math.min(after + batch, hi)
+      const { rows } = await db.rawQuery(
+        `SELECT id, name FROM releases WHERE id > ? AND id <= ? AND readable`,
+        [after, to]
+      )
+      looked += rows.length
+      const ids: number[] = []
+      for (const r of rows) {
+        if (!isUnreadableName(r.name)) continue
+        ids.push(Number(r.id))
+        if (sample.length < 40) sample.push(r.name)
+      }
+      changed += ids.length
+      if (this.apply) {
+        if (ids.length > 0) {
+          await db.rawQuery(`UPDATE releases SET readable = false WHERE id = ANY(?::bigint[])`, [
+            ids,
+          ])
+          // Their probe candidates to the unreadable tiers, which the prober takes first (LOG.md (130)).
+          await db.rawQuery(
+            `UPDATE probe_candidates SET unreadable = true WHERE release_id = ANY(?::bigint[]) AND NOT unreadable`,
+            [ids]
+          )
+        }
+        await db.rawQuery(
+          `INSERT INTO worker_cursors (name, at, key) VALUES (?, now(), ?) ON CONFLICT (name) DO UPDATE SET at = excluded.at, key = excluded.key`,
+          [CURSOR, String(to)]
+        )
+      }
+      after = to
+      if (Math.floor(after / 2_000_000) !== Math.floor((after - batch) / 2_000_000)) {
+        this.logger.info(
+          `Up to id ${after} of ${hi}: ${changed} unreadable of ${looked} readable (${Math.round((Date.now() - started) / 1000)} s).`
+        )
+      }
+    }
+    this.logger.info(
+      `${this.apply ? 'Cleared' : 'Would clear'} ${changed} of ${looked} readable rows.`
+    )
+    if (!this.apply) for (const n of sample) this.logger.info(`  ${n}`)
+  }
+}

commands/releases_regroup.ts

@@ -3,6 +3,7 @@ import type { CommandOptions } from '@adonisjs/core/types/ace'
 import db from '@adonisjs/lucid/services/db'
 import { decodeChunk } from '#lib/chunks/index'
 import { releaseIdentity } from '#lib/release/index'
+import { sourceCode } from '#lib/release/source'
 
 /**
  * What lib/release's identity makes of releases already in the catalog: a
@@ -27,7 +28,7 @@ export default class ReleasesRegroup extends BaseCommand {
   async run() {
     for (const name of this.group ?? []) {
       const { rows } = await db.rawQuery(
-        `SELECT DISTINCT c.release_id FROM release_chunks c JOIN ranges r ON r.id = c.range_id JOIN groups g ON g.id = r.group_id WHERE g.name = ?`,
+        `SELECT DISTINCT c.release_id FROM release_chunks c JOIN blocks b ON b.id = c.block_id JOIN groups g ON g.id = b.group_id WHERE g.name = ?`,
         [name]
       )
       const ids = rows.map((r: { release_id: string }) => r.release_id)
@@ -54,7 +55,7 @@ export default class ReleasesRegroup extends BaseCommand {
     }
     if (this.show?.length) return
     for (const source of ['subject', 'post', 'stem']) {
-      const rows = await db.from('releases').where('source', source).where('files', '>', 1).orderByRaw('random()').limit(this.sample).select('id', 'name', 'files')
+      const rows = await db.from('releases').where('source', sourceCode(source)).where('files', '>', 1).orderByRaw('random()').limit(this.sample).select('id', 'name', 'files')
       let split = 0
       let pieces = 0
       const examples: string[] = []

commands/replay_corpus.ts

@@ -0,0 +1,396 @@
+import { BaseCommand, flags } from '@adonisjs/core/ace'
+import type { CommandOptions } from '@adonisjs/core/types/ace'
+import { spawn } from 'node:child_process'
+import { access, readdir, readFile, writeFile } from 'node:fs/promises'
+import { tmpdir } from 'node:os'
+import { join, resolve } from 'node:path'
+import db from '@adonisjs/lucid/services/db'
+import coordinatorConfig from '#config/coordinator'
+import BatchStore from '#services/batch_store'
+import Leases from '#services/leases'
+import RuleSets from '#services/rule_sets'
+import { fromWire, type WireBatch } from '#lib/edge/index'
+
+/**
+ * Only a database named like one of these is written: the replay leases and finishes ranges and fills the catalog.
+ * Not the test database: the rows left behind failed 27 functional tests (LOG.md 2026-10-26 (111)). With
+ * REPLAY_INTO_CATALOG=1 any database but the test one: voyager 3 deployed, history replayed from the corpus into
+ * the catalog itself (deploy/replay-window.sh, LOG.md (116)).
+ */
+const SCRATCH_DB = /_(replay|scratch)$/
+const TEST_DB = /_test$/
+const NODE_NAME = 'replay'
+
+interface StoredRange {
+  id: number
+  start: number
+  end: number
+}
+
+/**
+ * The replay harness (REBUILD.md, "Voyager 3", 8): a group's stored corpus ranges in article order through one
+ * window in the Rust edge (`edge-bench --group`, the window carried from range to range) and the batch store into a
+ * scratch catalog, as the worker would store them: each range leased to a node named "replay", its batches stored
+ * at its resume point, the range finished when its last batch is in. Then a report: releases finished, complete,
+ * partial and provisional, and the catalog's bytes per header (the NVMe tables apart from those on CHUNK_TABLESPACE).
+ *
+ *   node ace replay:corpus --group alt.binaries.ath [--dir CORPUS_DIR] [--from A] [--to B] [--limit N] [--batch-chunks 5]
+ *
+ * Refuses a database not named *_replay or *_scratch. The rules are the catalog's current rule set
+ * (rule_sets, the groups' priorities): copy `groups` and `rule_sets` from the catalog for a faithful replay.
+ */
+export default class ReplayCorpus extends BaseCommand {
+  static commandName = 'replay:corpus'
+  static description =
+    "Replays a group's corpus ranges through the Rust edge and the batch store into a scratch catalog"
+  static options: CommandOptions = { startApp: true }
+
+  @flags.string({ description: 'The group (its directory under the corpus)', required: true })
+  declare group: string
+
+  @flags.string({ description: 'The corpus directory (default: CORPUS_DIR)' })
+  declare dir?: string
+
+  @flags.number({ description: 'Replay this many ranges at most (default: all stored)' })
+  declare limit?: number
+
+  @flags.number({ description: 'Skip ranges ending before this article (continue a replay)' })
+  declare from?: number
+
+  @flags.number({
+    description: "Skip ranges starting after this article (one window of the group's history)",
+  })
+  declare to?: number
+
+  @flags.number({
+    description: 'Chunks of 100k articles per batch, as the worker posts them',
+    default: 5,
+  })
+  declare batchChunks: number
+
+  @flags.string({
+    description:
+      'The edge-bench binary (default: EDGE_BENCH, edge/target/release/edge-bench, /usr/local/bin/edge-bench)',
+  })
+  declare bench?: string
+
+  @flags.boolean({
+    description:
+      'Leave the window open at the end (the last ranges provisional), to continue later',
+  })
+  declare noFinish: boolean
+
+  async run() {
+    const database = db.connection().getReadClient().client.config.connection.database as string
+    const intoCatalog = process.env.REPLAY_INTO_CATALOG === '1' && !TEST_DB.test(database)
+    if (!SCRATCH_DB.test(database) && !intoCatalog) {
+      this.logger.error(
+        `The replay writes the catalog: it only runs against a database named *_replay or *_scratch (this is "${database}"; REPLAY_INTO_CATALOG=1 allows it)`
+      )
+      this.exitCode = 1
+      return
+    }
+    const dir = this.dir ?? coordinatorConfig.corpus.dir
+    if (!dir) {
+      this.logger.error('No corpus directory: --dir or CORPUS_DIR')
+      this.exitCode = 1
+      return
+    }
+    const bench = await this.benchBinary()
+    if (!bench) {
+      this.logger.error(
+        'No edge-bench binary found: build it (cargo build --release --manifest-path edge/Cargo.toml) or set EDGE_BENCH'
+      )
+      this.exitCode = 1
+      return
+    }
+    const groupDir = join(dir, this.group)
+    const stored = await this.storedRanges(groupDir)
+    if (stored.length === 0) {
+      this.logger.error(`${groupDir}: no complete range`)
+      this.exitCode = 1
+      return
+    }
+    const nodeId = await this.nodeId()
+    const groupId = await this.groupId()
+    const ranges = await this.leaseRanges(groupId, nodeId, stored)
+    const rules = await RuleSets.current()
+    const rulesPath = join(tmpdir(), `replay-rules-${process.pid}.json`)
+    await writeFile(rulesPath, JSON.stringify(rules))
+    this.logger.info(
+      `${this.group}: ${ranges.length} ranges (${ranges[0].start}–${ranges[ranges.length - 1].end}) through ${bench}, rules v${rules.version}, into ${database}.`
+    )
+
+    // edge-bench walks the group directory in article order itself; --from, --to and --limit cut its list as ours.
+    const args = [
+      '--group',
+      groupDir,
+      '--rules',
+      rulesPath,
+      '--batch-chunks',
+      String(this.batchChunks),
+      '--wire',
+      '-',
+    ]
+    if (this.from) args.push('--from', String(this.from))
+    if (this.to) args.push('--to', String(this.to))
+    if (this.limit) args.push('--limit', String(this.limit))
+    if (this.noFinish) args.push('--no-finish')
+    const started = Date.now()
+    const child = spawn(bench, args, { stdio: ['ignore', 'pipe', 'inherit'] })
+    const totals = {
+      batches: 0,
+      headers: 0,
+      releases: 0,
+      newReleases: 0,
+      finalized: 0,
+      candidates: 0,
+      storeMs: 0,
+    }
+    let current: StoredRange | null = null
+    const storeLine = async (line: string) => {
+      const batch = fromWire(JSON.parse(line) as WireBatch)
+      if (!current || batch.from < current.start || batch.from > current.end) {
+        current = ranges.find((r) => batch.from >= r.start && batch.from <= r.end) ?? null
+        if (!current) throw new Error(`Batch at ${batch.from} is in no leased range`)
+      }
+      const at = Date.now()
+      const result = await BatchStore.store(current.id, nodeId, batch)
+      totals.storeMs += Date.now() - at
+      totals.batches++
+      totals.headers += batch.headers
+      totals.releases += result.releases
+      totals.newReleases += result.newReleases
+      totals.finalized += result.finalized
+      totals.candidates += result.candidates
+      if (batch.next > current.end) {
+        await Leases.finish(current.id, nodeId)
+        const seconds = (Date.now() - started) / 1000
+        this.logger.info(
+          `  range ${current.start}–${current.end} done: ${totals.headers.toLocaleString('en')} headers so far in ${seconds.toFixed(0)} s ` +
+            `(${Math.round(totals.headers / seconds).toLocaleString('en')}/s), ${totals.releases} releases stored (${totals.finalized} finished later), store ${(totals.storeMs / 1000).toFixed(0)} s`
+        )
+      }
+    }
+    // Lines pulled as the store keeps up: the child's stdout only moves when a chunk is read, so batches don't pile up.
+    let rest = ''
+    let failure: Error | null = null
+    try {
+      for await (const chunk of child.stdout as AsyncIterable<Buffer>) {
+        rest += chunk.toString('utf8')
+        let nl: number
+        while ((nl = rest.indexOf('\n')) >= 0) {
+          const line = rest.slice(0, nl)
+          rest = rest.slice(nl + 1)
+          if (line.length > 0) await storeLine(line)
+        }
+      }
+      if (rest.trim().length > 0) await storeLine(rest)
+    } catch (error: any) {
+      failure = error
+      child.kill('SIGTERM')
+    }
+    const code = await new Promise<number | null>((done) =>
+      child.exitCode !== null ? done(child.exitCode) : child.on('exit', done)
+    )
+    if (failure) {
+      this.logger.error(`Replay failed: ${failure.message}`)
+      this.exitCode = 1
+      return
+    }
+    if (code !== 0) {
+      this.logger.error(`edge-bench exited with ${code}`)
+      this.exitCode = 1
+      return
+    }
+    await this.report(groupId, ranges, Date.now() - started)
+  }
+
+  private async benchBinary(): Promise<string | null> {
+    const candidates = [
+      this.bench,
+      process.env.EDGE_BENCH,
+      resolve('edge/target/release/edge-bench'),
+      '/usr/local/bin/edge-bench',
+    ].filter((p): p is string => !!p)
+    for (const path of candidates) {
+      if (
+        await access(path).then(
+          () => true,
+          () => false
+        )
+      )
+        return path
+    }
+    return null
+  }
+
+  /** The group directory's complete ranges (a range.json), in article order, cut by --from, --to and --limit. */
+  private async storedRanges(groupDir: string): Promise<{ start: number; end: number }[]> {
+    const names = await readdir(groupDir).catch(() => [] as string[])
+    const ranges: { start: number; end: number }[] = []
+    for (const name of names) {
+      const m = /^(\d+)-(\d+)$/.exec(name)
+      if (!m) continue
+      const json = await readFile(join(groupDir, name, 'range.json'), 'utf8').catch(() => null)
+      if (!json) continue
+      ranges.push({ start: Number(m[1]), end: Number(m[2]) })
+    }
+    ranges.sort((a, b) => a.start - b.start)
+    let chosen = this.from ? ranges.filter((r) => r.end >= this.from!) : ranges
+    if (this.to) chosen = chosen.filter((r) => r.start <= this.to!)
+    if (this.limit) chosen = chosen.slice(0, this.limit)
+    return chosen
+  }
+
+  private async nodeId(): Promise<number> {
+    const row = await db.from('nodes').where('name', NODE_NAME).first()
+    if (row) return Number(row.id)
+    const [inserted] = await db.table('nodes').insert({ name: NODE_NAME }).returning('id')
+    return Number(inserted.id)
+  }
+
+  /** The group's row (inserted after the last priority when the scratch catalog doesn't have it). */
+  private async groupId(): Promise<number> {
+    const row = await db.from('groups').where('name', this.group).first()
+    if (row) return Number(row.id)
+    const { rows } = await db.rawQuery(`SELECT coalesce(max(priority), 0) + 1 AS next FROM groups`)
+    const [inserted] = await db
+      .table('groups')
+      .insert({ name: this.group, priority: Number(rows[0].next), active: true })
+      .returning('id')
+    this.logger.info(`Group ${this.group} added with priority ${rows[0].next}.`)
+    return Number(inserted.id)
+  }
+
+  /** The ranges as leased to the replay node, from their start (an earlier replay's rows are taken over). */
+  private async leaseRanges(
+    groupId: number,
+    nodeId: number,
+    stored: { start: number; end: number }[]
+  ): Promise<StoredRange[]> {
+    const out: StoredRange[] = []
+    const now = new Date()
+    for (const r of stored) {
+      const lease = {
+        status: 'leased',
+        node_id: nodeId,
+        next_id: r.start,
+        headers: 0,
+        posted_low: null,
+        posted_high: null,
+        leased_at: now,
+        heartbeat_at: now,
+        done_at: null,
+        error: null,
+        capture: true,
+      }
+      const existing = await db
+        .from('ranges')
+        .where('group_id', groupId)
+        .where('start_id', r.start)
+        .first()
+      if (existing) {
+        await db
+          .from('ranges')
+          .where('id', existing.id)
+          .update({ ...lease, end_id: r.end, attempts: Number(existing.attempts) + 1 })
+        out.push({ id: Number(existing.id), start: r.start, end: r.end })
+        continue
+      }
+      const [inserted] = await db
+        .table('ranges')
+        .insert({
+          ...lease,
+          group_id: groupId,
+          start_id: r.start,
+          end_id: r.end,
+          priority: 0,
+          attempts: 1,
+        })
+        .returning('id')
+      out.push({ id: Number(inserted.id), start: r.start, end: r.end })
+    }
+    return out
+  }
+
+  /** What the replay left in the catalog (this group's releases; the sizes are the whole catalog's), per article. */
+  private async report(groupId: number, ranges: StoredRange[], elapsedMs: number) {
+    const ids = ranges.map((r) => r.id)
+    const { rows: counts } = await db.rawQuery(
+      `SELECT count(*)::bigint AS releases,
+              count(settled_at)::bigint AS finished,
+              count(*) FILTER (WHERE settled_at IS NULL)::bigint AS provisional,
+              count(*) FILTER (WHERE edge)::bigint AS edges,
+              count(*) FILTER (WHERE complete)::bigint AS complete,
+              count(*) FILTER (WHERE complete = false)::bigint AS incomplete,
+              count(*) FILTER (WHERE complete IS NULL)::bigint AS undecided,
+              count(*) FILTER (WHERE readable)::bigint AS readable,
+              count(*) FILTER (WHERE has_par2)::bigint AS with_par2,
+              coalesce(sum(files), 0)::bigint AS files,
+              coalesce(sum(segments), 0)::bigint AS segments,
+              coalesce(sum(bytes), 0)::bigint AS bytes
+       FROM releases WHERE group_id = ?`,
+      [groupId]
+    )
+    const { rows: sources } = await db.rawQuery(
+      `SELECT source, count(*)::bigint AS n FROM releases WHERE group_id = ? GROUP BY source ORDER BY source`,
+      [groupId]
+    )
+    const { rows: pending } = await db.rawQuery(
+      `SELECT count(*)::bigint AS n, coalesce(sum(segments), 0)::bigint AS segments, count(*) FILTER (WHERE jsonb_array_length(candidates) = 0)::bigint AS unprobable
+       FROM pending_releases WHERE group_id = ?`,
+      [groupId]
+    )
+    const { rows: headerRows } = await db.rawQuery(
+      `SELECT coalesce(sum(headers), 0)::bigint AS headers FROM ranges WHERE id = ANY(?::bigint[])`,
+      [ids]
+    )
+    const headers = Number(headerRows[0].headers)
+    const { rows: sizes } = await db.rawQuery(
+      `SELECT CASE WHEN c.relname LIKE 'release_chunks%' THEN 'release_chunks' ELSE c.relname END AS name,
+              sum(pg_total_relation_size(c.oid))::bigint AS bytes
+       FROM pg_class c JOIN pg_namespace n ON n.oid = c.relnamespace
+       WHERE n.nspname = 'public' AND c.relkind = 'r'
+         AND (c.relname IN ('releases', 'release_texts', 'posters', 'group_sets', 'probe_candidates', 'rule_counts', 'rule_samples', 'blocks', 'pending_releases')
+              OR c.relname LIKE 'release_chunks%')
+       GROUP BY 1 ORDER BY 2 DESC`
+    )
+    // On CHUNK_TABLESPACE (the HDD) in the catalog: the chunks (migration 3200), inspections and candidates (6000).
+    const hdd = new Set([
+      'release_chunks',
+      'probe_candidates',
+      'release_inspections',
+      'pending_releases',
+    ])
+    let nvme = 0
+    let other = 0
+    const c = counts[0]
+    const fmt = (n: unknown) => Number(n).toLocaleString('en')
+    this.logger.info(
+      `Replay of ${ranges.length} ranges: ${fmt(headers)} headers in ${(elapsedMs / 1000).toFixed(0)} s (${fmt(Math.round(headers / Math.max(1, elapsedMs / 1000)))}/s).`
+    )
+    this.logger.info(
+      `Releases ${fmt(c.releases)}: finished ${fmt(c.finished)}, provisional ${fmt(c.provisional)}, edges ${fmt(c.edges)}; ` +
+        `complete ${fmt(c.complete)}, incomplete ${fmt(c.incomplete)}, undecided ${fmt(c.undecided)}; readable ${fmt(c.readable)}, with par2 ${fmt(c.with_par2)}; ` +
+        `${fmt(c.files)} files, ${fmt(c.segments)} segments (${((Number(c.segments) / Math.max(1, headers)) * 100).toFixed(1)}% of the headers), ${(Number(c.bytes) / 1e12).toFixed(2)} TB.`
+    )
+    this.logger.info(
+      `By source: ${sources.map((s: any) => `${s.source}: ${fmt(s.n)}`).join(', ')}.`
+    )
+    this.logger.info(
+      `Pending (unnamed, outside the catalog): ${fmt(pending[0].n)} releases, ${fmt(pending[0].segments)} segments, ${fmt(pending[0].unprobable)} without a probe target.`
+    )
+    for (const s of sizes) {
+      const bytes = Number(s.bytes)
+      if (hdd.has(s.name)) other += bytes
+      else nvme += bytes
+      this.logger.info(
+        `  ${s.name}: ${(bytes / 1e6).toFixed(1)} MB, ${(bytes / Math.max(1, headers)).toFixed(2)} B/header`
+      )
+    }
+    this.logger.info(
+      `NVMe tables: ${(nvme / 1e6).toFixed(1)} MB, ${(nvme / Math.max(1, headers)).toFixed(2)} B/header; HDD tables (CHUNK_TABLESPACE): ${(other / 1e6).toFixed(1)} MB, ${(other / Math.max(1, headers)).toFixed(2)} B/header (the whole catalog's sizes over this group's headers: only the first group's are its own).`
+    )
+  }
+}

commands/rules_classes.ts

@@ -0,0 +1,54 @@
+import { BaseCommand, flags } from '@adonisjs/core/ace'
+import type { CommandOptions } from '@adonisjs/core/types/ace'
+import db from '@adonisjs/lucid/services/db'
+import RuleSets from '#services/rule_sets'
+import { classKey, describeClass } from '#lib/release/index'
+
+/**
+ * Publishes a rule set whose dropped classes are the pending ladder's (pending_classes.dropped; LOG.md (130)): the
+ * edge keeps unnamed releases of those classes out of the pending store from then on, their explored sample
+ * excepted. Run again as the record grows; unchanged classes publish nothing.
+ */
+export default class RulesClasses extends BaseCommand {
+  static commandName = 'rules:classes'
+  static description = "Publishes a rule set with the pending ladder's dropped classes"
+  static options: CommandOptions = { startApp: true }
+
+  @flags.boolean({ description: 'Show what would be published, write nothing' })
+  declare dryRun: boolean
+
+  async run() {
+    const { rows } = await db.rawQuery(
+      `SELECT c.tool, g.name AS grp, c.klass, c.probed, c.named FROM pending_classes c JOIN groups g ON g.id = c.group_id WHERE c.dropped ORDER BY 1, 2, 3`
+    )
+    const next = new Set<string>(rows.map((r: any) => classKey(r.tool, r.grp, Number(r.klass))))
+    const { version, priorities: _p, ...current } = await RuleSets.current()
+    const before = new Set(current.droppedClasses ?? [])
+    const added = [...next].filter((c) => !before.has(c))
+    const removed = [...before].filter((c) => !next.has(c))
+    if (added.length === 0 && removed.length === 0) {
+      this.logger.info(
+        `Unchanged: rule set ${version} has the ${before.size} dropped classes already.`
+      )
+      return
+    }
+    for (const r of rows.slice(0, 40)) {
+      this.logger.info(
+        `  ${r.tool.padEnd(10)} ${r.grp.padEnd(32)} ${describeClass(Number(r.klass)).padEnd(26)} ${r.named} of ${r.probed} named`
+      )
+    }
+    if (this.dryRun) {
+      this.logger.info(
+        `Would publish ${next.size} dropped classes (${added.length} added, ${removed.length} removed).`
+      )
+      return
+    }
+    const published = await RuleSets.publish(
+      { ...current, droppedClasses: [...next].sort() },
+      `${next.size} dropped classes from the pending ladder`
+    )
+    this.logger.info(
+      `Rule set ${published}: ${next.size} dropped classes (${added.length} added, ${removed.length} removed).`
+    )
+  }
+}

config/coordinator.ts

@@ -31,6 +31,22 @@ export default {
   },
   /** A leased range without a heartbeat for this long goes back to the queue. */
   leaseTtlSeconds: env.get('LEASE_TTL_SECONDS', 15 * 60),
+  /**
+   * The ranges group leases hand out (REBUILD.md design 1), by priority: the tail is negative, the capture windows
+   * 0–2 (2016, 2021, 2025), the history by its position in the group (0 oldest … 1000 newest: its first buckets
+   * share 0–2 with the windows) or 4000+. Bounds so a deployment can run one window first, with
+   * LEASE_CAPTURE_ONLY so the history's 0 isn't taken along (homelab 2026-10-26: 0..0 and capture only, the 2016
+   * window; LOG.md (118)); unbounded and any range by default.
+   */
+  leasePriorityMin: env.get('LEASE_MIN_PRIORITY', -1_000_000),
+  leasePriorityMax: env.get('LEASE_MAX_PRIORITY', 1_000_000),
+  leaseCaptureOnly: env.get('LEASE_CAPTURE_ONLY', false),
+  /**
+   * How long a range's stream waits for a chunk the prefetcher hasn't spooled yet (app/controllers/node_controller.ts).
+   * Long: a chunk that keeps coming in isn't given up (only a stalled one is), and some groups' chunks take 20+
+   * minutes (a.b.0day.stuffz, 17 MB in 22 min; LOG.md 2026-10-10 (3)).
+   */
+  streamChunkWaitMs: env.get('STREAM_CHUNK_WAIT_MINUTES', 60) * 60 * 1000,
   /** NNTP connections the coordinator fetches headers with: at most (its share, nntpShares), and at least while sharing. */
   nntpConnections: env.get('NNTP_CONNECTIONS', 4),
   nntpMinConnections: env.get('NNTP_MIN_CONNECTIONS', 8),
@@ -103,6 +119,14 @@ export default {
     batch: 400,
     /** BODY commands in flight per connection (pipelined; XoverClient.bodies). */
     pipeline: env.get('PROBE_PIPELINE', 8),
+    /**
+     * The pending ladder (REBUILD.md decision f; app/services/pending_ladder.ts): the share of each claim taken
+     * from pending_releases at least (the catalog's own candidates come first, the rest of the claim is theirs).
+     */
+    pendingShare: env.get('PENDING_SHARE', 0.25),
+    /** A signature (tool, group, poster) with this many climbs and at most pendingMaxNamed of them named is dropped. */
+    pendingMinProbes: env.get('PENDING_MIN_PROBES', 100),
+    pendingMaxNamed: env.get('PENDING_MAX_NAMED', 0.005),
   },
   /**
    * The catalog's tablespace may grow to this many GB (app/services/catalog_space.ts); above it only the tail and
@@ -138,6 +162,11 @@ export default {
      */
     filter: env.get('CORPUS_FILTER', true),
   },
+  /**
+   * The group windows' snapshots (REBUILD.md, "Voyager 3: window indexing", 5; app/services/group_leases.ts): a
+   * file a group on bulk, next to the spool by default.
+   */
+  snapshotDir: env.get('SNAPSHOT_DIR', `${env.get('SPOOL_DIR', app.tmpPath('spool'))}/snapshots`),
   spool: {
     dir: env.get('SPOOL_DIR', app.tmpPath('spool')),
     /** Ranges kept fetched ahead, per active lease. */

database/migrations/1770000006000_voyager3_catalog.ts

@@ -0,0 +1,177 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+import env from '#start/env'
+
+/**
+ * Voyager 3's catalog (REBUILD.md, "Voyager 3: window indexing", 4 and 7), built from a reset, not converted in
+ * place. This migration IS the reset: it empties every table the indexing recomputes (releases, chunks,
+ * inspections, candidates, par2 data, file hashes, verdicts, merges, rule counts and samples; what survives:
+ * groups, ranges, the corpus, probe_cache, spots, nntp_shares, nodes, users), so it refuses to run on a catalog
+ * with releases unless VOYAGER3_RESET=1 (the user's go-ahead; resets are paused since 2026-10-12). Then:
+ * - posters get an id; group_sets holds each distinct group combination once (~140k); releases is created anew
+ *   without its text copies (poster_id, group_set_id, the keys as 20-byte bytea, source a smallint), 8-byte
+ *   columns first, with `edge` (still open at its block's end, for the block merger) and settled_at as the finish
+ *   (null: a tail release still provisional). The category and judge triggers follow the new columns;
+ *   releases_category_newest_idx (5.3 GB, used once) goes, and the trigram indexes: search matches release_texts
+ *   (migration 6200), which holds the name and every searchable string learned later.
+ * - fragments and range_blocks go with the fragment merger and the settler (the window finishes a release once);
+ *   release_final() goes: a release is final once finished.
+ * - release_chunks, rule_counts and rule_samples are keyed by the block that wrote them (block_id; migration 6100).
+ * - release_inspections and probe_candidates move to CHUNK_TABLESPACE (the HDD) when set: written once, read per
+ *   release, not by search.
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.defer(async (db) => {
+      const { rows } = await db.rawQuery(`SELECT count(*)::int AS n FROM releases`)
+      if (rows[0].n > 0 && process.env.VOYAGER3_RESET !== '1') {
+        throw new Error(
+          `The voyager 3 catalog migration empties the catalog (${rows[0].n} releases): run it with VOYAGER3_RESET=1 after the go-ahead`
+        )
+      }
+    })
+    this.schema.raw(`
+      TRUNCATE releases, release_chunks, release_inspections, probe_candidates, file_hashes, par2_files, par2_sets,
+        release_verdicts, release_merges, release_spots, release_passwords, rule_counts, rule_samples, judge_dirty,
+        judge_full, fold_candidates, posters, worker_cursors RESTART IDENTITY`)
+    // Spots matched to releases that are gone: the spot inspector matches them again.
+    this.schema.raw(`UPDATE spots SET release_id = NULL WHERE release_id IS NOT NULL`)
+    this.schema.raw(`UPDATE spot_files SET release_id = NULL WHERE release_id IS NOT NULL`)
+
+    this.schema.raw(`DROP TABLE IF EXISTS fragments`)
+    this.schema.raw(`DROP TABLE IF EXISTS range_blocks`)
+    this.schema.raw(
+      `DROP FUNCTION IF EXISTS release_final(integer, bigint, bigint, timestamptz, timestamptz, boolean)`
+    )
+    this.schema.raw(
+      `DROP FUNCTION IF EXISTS release_final(integer, bigint, bigint, timestamptz, timestamptz)`
+    )
+
+    // Posters by id (a release's poster_id); searched by trigram as before (poster_search).
+    this.schema.raw(`DROP TABLE IF EXISTS posters`)
+    this.schema.raw(`
+      CREATE TABLE posters (
+        id serial PRIMARY KEY,
+        poster text NOT NULL UNIQUE
+      )`)
+    this.schema.raw(`CREATE INDEX posters_trgm ON posters USING gin (poster gin_trgm_ops)`)
+    // Each distinct combination of groups a release was posted to, once: a sorted jsonb array of names.
+    this.schema.raw(`
+      CREATE TABLE group_sets (
+        id serial PRIMARY KEY,
+        groups jsonb NOT NULL UNIQUE
+      )`)
+
+    this.schema.raw(`DROP TABLE IF EXISTS releases CASCADE`)
+    this.schema.raw(`
+      CREATE TABLE releases (
+        id bigserial PRIMARY KEY,
+        -- The upload's posting window (first and last article's Date).
+        posted_at timestamptz,
+        last_posted_at timestamptz,
+        first_article bigint,
+        last_article bigint,
+        segments bigint NOT NULL DEFAULT 0,
+        bytes bigint NOT NULL DEFAULT 0,
+        -- The finish: the window closed on it (or the block ended); null while a tail release is provisional.
+        settled_at timestamptz,
+        inspected_at timestamptz,
+        first_seen_at timestamptz NOT NULL DEFAULT now(),
+        -- The owner group (null: a spot's release), the set of groups it was posted to, its poster.
+        group_id integer,
+        group_set_id integer NOT NULL,
+        poster_id integer,
+        file_total integer,
+        files integer NOT NULL DEFAULT 0,
+        chunks integer NOT NULL DEFAULT 0,
+        category integer,
+        -- lib/release/source.ts: 1 subject, 2 post, 3 stem, 4 spot, 5 nzb, 6 album.
+        source smallint NOT NULL,
+        readable boolean NOT NULL DEFAULT false,
+        has_par2 boolean NOT NULL DEFAULT false,
+        complete boolean,
+        passworded boolean,
+        sidecar_only boolean,
+        -- Still open when its block ended: the block merger joins it with the neighbour block's part by identity.
+        edge boolean NOT NULL DEFAULT false,
+        -- SHA-1s (lib/release/identity.ts): the identity (poster, name, group), and the release (identity and upload).
+        release_key bytea NOT NULL,
+        identity_key bytea NOT NULL,
+        name text NOT NULL,
+        inner_name varchar(255)
+      )`)
+    this.schema.raw(`CREATE UNIQUE INDEX releases_release_key_unique ON releases (release_key)`)
+    this.schema.raw(`CREATE INDEX releases_identity_idx ON releases (identity_key, group_id)`)
+    // No trigram indexes here: search matches release_texts (migration 6200), where the name is stored once more.
+    this.schema.raw(
+      `CREATE INDEX releases_newest_idx ON releases (posted_at DESC NULLS LAST, id DESC)`
+    )
+    this.schema.raw(
+      `CREATE INDEX releases_poster_newest_idx ON releases (poster_id, posted_at DESC NULLS LAST, id DESC)`
+    )
+    this.schema.raw(`CREATE INDEX releases_passworded_idx ON releases (id) WHERE passworded`)
+    this.schema.raw(
+      `CREATE INDEX releases_edge_idx ON releases (group_id, identity_key) WHERE edge`
+    )
+    this.schema.raw(
+      `CREATE INDEX releases_provisional_idx ON releases (id) WHERE settled_at IS NULL`
+    )
+
+    // The category from the name, the inner name, or the groups (migration 5200), the groups now in group_sets.
+    this.schema.raw(`
+      CREATE OR REPLACE FUNCTION releases_set_category() RETURNS trigger LANGUAGE plpgsql AS $$
+      BEGIN
+        NEW.category := release_category_for(NEW.name, NEW.inner_name, (SELECT groups FROM group_sets WHERE id = NEW.group_set_id),
+                                             CASE WHEN TG_OP = 'UPDATE' THEN OLD.category ELSE NEW.category END);
+        RETURN NEW;
+      END $$`)
+    this.schema.raw(`
+      CREATE TRIGGER releases_category BEFORE INSERT OR UPDATE OF name, inner_name, group_set_id ON releases
+      FOR EACH ROW EXECUTE FUNCTION releases_set_category()`)
+    // The judge's dirty marks (migration 4900), on the columns its rules read.
+    this.schema.raw(`
+      CREATE TRIGGER judge_dirty_insert AFTER INSERT ON releases REFERENCING NEW TABLE AS new_rows
+      FOR EACH STATEMENT EXECUTE FUNCTION judge_dirty_releases()`)
+    this.schema.raw(`
+      CREATE TRIGGER judge_dirty_delete AFTER DELETE ON releases REFERENCING OLD TABLE AS old_rows
+      FOR EACH STATEMENT EXECUTE FUNCTION judge_dirty_releases()`)
+    this.schema.raw(`
+      CREATE TRIGGER judge_dirty_update AFTER UPDATE ON releases FOR EACH ROW
+      WHEN (old.name IS DISTINCT FROM new.name OR old.poster_id IS DISTINCT FROM new.poster_id OR old.source IS DISTINCT FROM new.source
+        OR old.file_total IS DISTINCT FROM new.file_total OR old.passworded IS DISTINCT FROM new.passworded
+        OR old.settled_at IS DISTINCT FROM new.settled_at OR old.readable IS DISTINCT FROM new.readable
+        OR old.inner_name IS DISTINCT FROM new.inner_name OR old.group_id IS DISTINCT FROM new.group_id
+        OR old.first_article IS DISTINCT FROM new.first_article OR old.last_article IS DISTINCT FROM new.last_article
+        OR old.posted_at IS DISTINCT FROM new.posted_at OR old.last_posted_at IS DISTINCT FROM new.last_posted_at
+        OR old.complete IS DISTINCT FROM new.complete)
+      EXECUTE FUNCTION judge_dirty_releases()`)
+
+    // Chunks, rule counts and samples by the block that wrote them.
+    this.schema.raw(`ALTER TABLE release_chunks RENAME COLUMN range_id TO block_id`)
+    this.schema.raw(
+      `ALTER INDEX IF EXISTS release_chunks_range_idx RENAME TO release_chunks_block_idx`
+    )
+    this.schema.raw(`ALTER TABLE rule_counts RENAME COLUMN range_id TO block_id`)
+    this.schema.raw(`ALTER TABLE rule_samples RENAME COLUMN range_id TO block_id`)
+    this.schema.raw(
+      `ALTER INDEX IF EXISTS rule_samples_range_id_index RENAME TO rule_samples_block_id_index`
+    )
+
+    // Cold tables to the HDD (CHUNK_TABLESPACE), with their indexes; empty now, so at once.
+    const hdd = env.get('CHUNK_TABLESPACE')
+    if (hdd) {
+      for (const table of ['release_inspections', 'probe_candidates']) {
+        this.schema.raw(`ALTER TABLE ${table} SET TABLESPACE ${hdd}`)
+        this.schema.raw(`
+          DO $$ DECLARE i record; BEGIN
+            FOR i IN SELECT indexname FROM pg_indexes WHERE schemaname = 'public' AND tablename = '${table}' LOOP
+              EXECUTE format('ALTER INDEX %I SET TABLESPACE %I', i.indexname, '${hdd}');
+            END LOOP;
+          END $$`)
+      }
+    }
+  }
+
+  async down() {
+    throw new Error('The voyager 3 catalog migration cannot be undone (it emptied the catalog)')
+  }
+}

database/migrations/1770000006100_blocks.ts

@@ -0,0 +1,61 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * Blocks (REBUILD.md, "Voyager 3: window indexing", 1): what a worker leases. A block is a run of consecutive
+ * ranges of one group covering at least 30 days of posting time (planned from ranges.est_date and posted_low/high),
+ * read in article order through the 24 h window; blocks of a group run in parallel. Ranges stay the unit the
+ * prefetcher fetches (the spool, the corpus) and point at their block; the lease, resume point, heartbeat and
+ * counts move here. Part of the reset (migration 6000): every range is read again, from the corpus where stored.
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.schema.raw(`
+      CREATE TABLE blocks (
+        id bigserial PRIMARY KEY,
+        group_id integer NOT NULL REFERENCES groups (id) ON DELETE CASCADE,
+        -- The first range's start and the last range's end.
+        start_id bigint NOT NULL,
+        end_id bigint NOT NULL,
+        -- The resume point: the first article not stored yet (batches are refused out of order, as ranges' were).
+        next_id bigint NOT NULL,
+        -- The posting time the block covers: estimated when planned, the worker's measured span once done.
+        posted_low timestamptz,
+        posted_high timestamptz,
+        -- planned, leased, done, failed.
+        status varchar(8) NOT NULL DEFAULT 'planned',
+        -- As ranges': the tail before everything, then the history by era.
+        priority integer NOT NULL DEFAULT 0,
+        node_id integer,
+        attempts integer NOT NULL DEFAULT 0,
+        rule_version integer,
+        headers bigint NOT NULL DEFAULT 0,
+        releases integer NOT NULL DEFAULT 0,
+        leased_at timestamptz,
+        heartbeat_at timestamptz,
+        done_at timestamptz,
+        error text,
+        created_at timestamptz NOT NULL DEFAULT now(),
+        UNIQUE (group_id, start_id)
+      )`)
+    this.schema.raw(
+      `CREATE INDEX blocks_todo_idx ON blocks (priority, id) WHERE status IN ('planned', 'leased')`
+    )
+    this.schema.raw(`CREATE INDEX blocks_leased_idx ON blocks (leased_at) WHERE status = 'leased'`)
+    this.schema.raw(
+      `ALTER TABLE ranges ADD COLUMN IF NOT EXISTS block_id bigint REFERENCES blocks (id) ON DELETE SET NULL`
+    )
+    this.schema.raw(`CREATE INDEX IF NOT EXISTS ranges_block_idx ON ranges (block_id)`)
+    // The reset: every range planned again from its start (the corpus serves the stored ones). Priorities, capture,
+    // est_date, stored_at, plain_chunks and the measured posted_low/high stay.
+    this.schema.raw(`
+      UPDATE ranges SET status = 'planned', next_id = start_id, node_id = NULL, attempts = 0, spooled_chunks = 0,
+        spooled_at = NULL, leased_at = NULL, heartbeat_at = NULL, done_at = NULL, headers = 0, error = NULL,
+        rule_version = NULL
+      WHERE status <> 'planned' OR next_id <> start_id OR spooled_chunks <> 0 OR node_id IS NOT NULL`)
+  }
+
+  async down() {
+    this.schema.raw(`ALTER TABLE ranges DROP COLUMN IF EXISTS block_id`)
+    this.schema.raw(`DROP TABLE IF EXISTS blocks`)
+  }
+}

database/migrations/1770000006200_release_texts.ts

@@ -0,0 +1,35 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * release_texts (REBUILD.md, "Voyager 3 … Schema"): every string search matches against, one table, one trigram
+ * index. The release name goes in (kind 1) with whatever anyone learns later: the prober's file names inside an
+ * archive, NFO titles, par2 set names, media titles (lib/release/text_kind.ts). Appending is an insert, never an
+ * update of the wide releases row; a repeat insert is a no-op by the primary key; the judge's dirty mark follows
+ * (judge_dirty_by_release_id, migration 4900), so a rule can react to a name arriving later. On the NVMe: search
+ * reads it. releases keeps name and inner_name for display, without trigram indexes (migration 6000).
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.schema.raw(`
+      CREATE TABLE release_texts (
+        release_id bigint NOT NULL,
+        -- lib/release/text_kind.ts: 1 name, 2 file, 3 nfo, 4 par2, 5 media.
+        kind smallint NOT NULL,
+        text varchar(255) NOT NULL,
+        PRIMARY KEY (release_id, kind, text)
+      )`)
+    this.schema.raw(
+      `CREATE INDEX release_texts_trgm ON release_texts USING gin (text gin_trgm_ops) WITH (fastupdate = on, gin_pending_list_limit = 65536)`
+    )
+    this.schema.raw(
+      `CREATE TRIGGER judge_dirty_insert AFTER INSERT ON release_texts REFERENCING NEW TABLE AS new_rows FOR EACH STATEMENT EXECUTE FUNCTION judge_dirty_by_release_id()`
+    )
+    this.schema.raw(
+      `CREATE TRIGGER judge_dirty_delete AFTER DELETE ON release_texts REFERENCING OLD TABLE AS old_rows FOR EACH STATEMENT EXECUTE FUNCTION judge_dirty_by_release_id()`
+    )
+  }
+
+  async down() {
+    this.schema.raw(`DROP TABLE IF EXISTS release_texts`)
+  }
+}

database/migrations/1770000006300_pending_releases.ts

@@ -0,0 +1,60 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+import env from '#start/env'
+
+/**
+ * pending_releases (REBUILD.md decision f, 2026-10-26): nothing enters the catalog without a usable name. A release
+ * the window closed on whose files all have random names is written here instead of releases: its keys (as the
+ * catalog would store them), its counts and article span, the upload tool from its message-id (with the group its
+ * signature), and its probe targets by effort (the edge's candidates: a par2, an archive start, a lone file, a 7z
+ * tail), no chunk. The pending ladder probes them cheapest first and moves a named one into the catalog, reading its
+ * segments again from the corpus or the server; a signature that names nothing is discarded. On CHUNK_TABLESPACE
+ * (the HDD): written once, read by the ladder, never by search.
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.schema.raw(`
+      CREATE TABLE pending_releases (
+        -- lib/release/keys.ts: the identity, the group and the first article; the identity alone.
+        release_key bytea PRIMARY KEY,
+        identity_key bytea NOT NULL,
+        group_id integer NOT NULL,
+        -- The stem or post the edge keyed it by (a random name: no release name, but the inner-stem merge's handle).
+        name varchar(255) NOT NULL,
+        poster text,
+        posted_at timestamptz,
+        last_posted_at timestamptz,
+        first_article bigint NOT NULL,
+        last_article bigint NOT NULL,
+        files integer NOT NULL DEFAULT 0,
+        segments bigint NOT NULL DEFAULT 0,
+        bytes bigint NOT NULL DEFAULT 0,
+        -- A file misses parts.
+        partial boolean NOT NULL DEFAULT false,
+        -- False: a tail release still open; a later chunk of its key adds its counts up.
+        finished boolean NOT NULL DEFAULT true,
+        -- lib/rules uploadTool: ngPost, nyuu, hex32, random, other, …
+        tool varchar(16) NOT NULL,
+        -- [{kind, messageId, file}] by effort (lib/edge ProbeCandidate); empty: nothing to read, only a signature.
+        candidates jsonb NOT NULL DEFAULT '[]',
+        -- 0 waiting, 1 named (in the catalog), 2 discarded (its signature's verdict, or every probe gave nothing).
+        status smallint NOT NULL DEFAULT 0,
+        tries smallint NOT NULL DEFAULT 0,
+        created_at timestamptz NOT NULL DEFAULT now(),
+        probed_at timestamptz
+      )`)
+    // The ladder's order: a group's and tool's waiting releases, oldest first.
+    this.schema.raw(
+      `CREATE INDEX pending_releases_ladder_idx ON pending_releases (group_id, tool, status, created_at)`
+    )
+    const hdd = env.get('CHUNK_TABLESPACE')
+    if (hdd) {
+      this.schema.raw(`ALTER TABLE pending_releases SET TABLESPACE ${hdd}`)
+      this.schema.raw(`ALTER INDEX pending_releases_pkey SET TABLESPACE ${hdd}`)
+      this.schema.raw(`ALTER INDEX pending_releases_ladder_idx SET TABLESPACE ${hdd}`)
+    }
+  }
+
+  async down() {
+    this.schema.raw(`DROP TABLE IF EXISTS pending_releases`)
+  }
+}

database/migrations/1770000006400_pending_ladder.ts

@@ -0,0 +1,49 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * The pending ladder (REBUILD.md decision f, 2026-10-26): what the prober needs to work pending_releases cheapest
+ * first and to learn which signatures never name anything.
+ * - pending_releases: `file_total` (the counter's, for the catalog row once named), `inspections` (what each probe
+ *   showed, by kind), `named` (the usable name a probe found; status 4 until the namer moved it into the catalog).
+ *   Status: 0 waiting, 1 named and in the catalog, 2 discarded (its signature's verdict), 3 exhausted (every
+ *   target probed, no name), 4 named, awaiting the move, 9 taken by a running prober.
+ * - pending_signatures: per (tool, group, poster) how many pending releases were probed to the end and how many
+ *   got a name; `dropped` once probed ≥ PENDING_MIN_PROBES and named ≤ PENDING_MAX_NAMED of them (voyager 2's
+ *   garbage learner, LOG.md 2026-10-08 (5)). The poster '' row is the tool's and group's together.
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.schema.raw(`ALTER TABLE pending_releases ADD COLUMN IF NOT EXISTS file_total integer`)
+    this.schema.raw(
+      `ALTER TABLE pending_releases ADD COLUMN IF NOT EXISTS inspections jsonb NOT NULL DEFAULT '[]'`
+    )
+    this.schema.raw(`ALTER TABLE pending_releases ADD COLUMN IF NOT EXISTS named varchar(255)`)
+    // The ladder's claim: the oldest waiting ones; the namer's: the named ones.
+    this.schema.raw(
+      `CREATE INDEX IF NOT EXISTS pending_releases_waiting_idx ON pending_releases (created_at) WHERE status = 0`
+    )
+    this.schema.raw(
+      `CREATE INDEX IF NOT EXISTS pending_releases_naming_idx ON pending_releases (probed_at) WHERE status = 4`
+    )
+    this.schema.raw(`
+      CREATE TABLE pending_signatures (
+        tool varchar(16) NOT NULL,
+        group_id integer NOT NULL,
+        poster text NOT NULL DEFAULT '',
+        probed integer NOT NULL DEFAULT 0,
+        named integer NOT NULL DEFAULT 0,
+        dropped boolean NOT NULL DEFAULT false,
+        updated_at timestamptz NOT NULL DEFAULT now(),
+        PRIMARY KEY (tool, group_id, poster)
+      )`)
+  }
+
+  async down() {
+    this.schema.raw(`DROP TABLE IF EXISTS pending_signatures`)
+    this.schema.raw(`DROP INDEX IF EXISTS pending_releases_naming_idx`)
+    this.schema.raw(`DROP INDEX IF EXISTS pending_releases_waiting_idx`)
+    this.schema.raw(`ALTER TABLE pending_releases DROP COLUMN IF EXISTS named`)
+    this.schema.raw(`ALTER TABLE pending_releases DROP COLUMN IF EXISTS inspections`)
+    this.schema.raw(`ALTER TABLE pending_releases DROP COLUMN IF EXISTS file_total`)
+  }
+}

database/migrations/1770000006500_group_leases.ts

@@ -0,0 +1,33 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * Group leases (REBUILD.md, "Voyager 3: window indexing", 1 and 5; 2026-10-26): a worker thread holds a group, not
+ * a range, and indexes its ranges in article order through one carried window. One row a group: the holder and
+ * its heartbeat, the range it works, and the snapshot of its window (the article it stands at: the next one to
+ * index) that the coordinator keeps on bulk. Ranges stay what is planned, fetched and finished.
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.schema.raw(`
+      CREATE TABLE group_leases (
+        group_id integer PRIMARY KEY REFERENCES groups (id) ON DELETE CASCADE,
+        -- The holder; null between holders. A holder quiet for LEASE_TTL_SECONDS is taken over.
+        node_id integer,
+        leased_at timestamptz,
+        heartbeat_at timestamptz,
+        -- The range the holder works (null between ranges).
+        range_id bigint,
+        -- The window's snapshot: the article it stands at (a resume point), and when it was taken; null: none.
+        snapshot_next bigint,
+        snapshot_at timestamptz,
+        ranges_done integer NOT NULL DEFAULT 0,
+        headers bigint NOT NULL DEFAULT 0,
+        created_at timestamptz NOT NULL DEFAULT now()
+      )`)
+    this.schema.raw(`CREATE INDEX group_leases_held_idx ON group_leases (heartbeat_at) WHERE node_id IS NOT NULL`)
+  }
+
+  async down() {
+    this.schema.raw(`DROP TABLE IF EXISTS group_leases`)
+  }
+}

database/migrations/1770000006600_group_lease_lanes.ts

@@ -0,0 +1,27 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * Lease lanes (REBUILD.md, "Voyager 3: window indexing", 1, amended 2026-10-27): a group has two cursors, each with
+ * its own holder, window and snapshot. The `history` lane works the ranges the lease bounds hand out (the windows,
+ * the history), the `tail` lane the tail planner's (priority at or below the tail's): a thread on the tail never
+ * waits behind a group's history, and a held history group's tail isn't stuck behind its days of ranges.
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.schema.raw(
+      `ALTER TABLE group_leases ADD COLUMN lane varchar(8) NOT NULL DEFAULT 'history'`
+    )
+    this.schema.raw(`ALTER TABLE group_leases DROP CONSTRAINT group_leases_pkey`)
+    this.schema.raw(`ALTER TABLE group_leases ADD PRIMARY KEY (group_id, lane)`)
+    // Counts the leases handed out: odd and even ones start from the other half of the order's mix.
+    this.schema.raw(`CREATE SEQUENCE group_lease_turns`)
+  }
+
+  async down() {
+    this.schema.raw(`DROP SEQUENCE IF EXISTS group_lease_turns`)
+    this.schema.raw(`DELETE FROM group_leases WHERE lane <> 'history'`)
+    this.schema.raw(`ALTER TABLE group_leases DROP CONSTRAINT group_leases_pkey`)
+    this.schema.raw(`ALTER TABLE group_leases DROP COLUMN lane`)
+    this.schema.raw(`ALTER TABLE group_leases ADD PRIMARY KEY (group_id)`)
+  }
+}

database/migrations/1770000006700_ancient_ranges.ts

@@ -0,0 +1,20 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * One "ancient" range a group, articles 1 … (its first range − 1): what had expired on the server before the
+ * group was planned (the planner cuts from the low watermark). Status 'ancient': nothing to fetch, never leased,
+ * so the tiling is complete from article 1 and the span below the low watermark is on record (user, 2026-10-28).
+ * Planner.historyRows adds it for groups planned from now on.
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.schema.raw(
+      `INSERT INTO ranges (group_id, start_id, end_id, next_id, priority, status)
+       SELECT group_id, 1, min(start_id) - 1, 1, 0, 'ancient' FROM ranges GROUP BY group_id HAVING min(start_id) > 1`
+    )
+  }
+
+  async down() {
+    this.schema.raw(`DELETE FROM ranges WHERE status = 'ancient'`)
+  }
+}

database/migrations/1770000006800_pending_classes.ts

@@ -0,0 +1,57 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * The pending ladder by class (REBUILD.md decision f; LOG.md (130)): a pending release's class (lib/release/classes.ts:
+ * size bucket × name shape × partial, 1..70; 0 for rows from before) and the bytes of its first target, so the
+ * claim takes the classes with the best record first and the cheapest article within one; pending_classes per
+ * (tool, group, class) how many climbed to the end and how many got a name, `waiting` how many wait (kept by the
+ * batch store, the claim and the discard; recounted by pending:classes), `dropped` as pending_signatures.
+ * The SQL pending_class() is the TypeScript's pendingClass, for the batch store's insert and the backfill.
+ */
+export default class extends BaseSchema {
+  static disableTransactions = true
+
+  async up() {
+    this.schema.raw(`
+      CREATE OR REPLACE FUNCTION pending_class(name text, bytes bigint, partial boolean) RETURNS smallint LANGUAGE sql IMMUTABLE AS $$
+        SELECT ((CASE WHEN bytes < 1000000 THEN 0 WHEN bytes < 10000000 THEN 1 WHEN bytes < 100000000 THEN 2 WHEN bytes < 1000000000 THEN 3 ELSE 4 END) * 14
+          + (CASE WHEN s ~ '^[0-9a-fA-F]{16,}$' THEN 1
+                  WHEN s ~ '^[0-9]{6,}$' THEN 2
+                  WHEN s ~ '^[a-z2-7]{20,}$' AND s ~ '[2-7]' THEN 3
+                  WHEN s ~ '^[A-Za-z]+$' THEN 4
+                  WHEN s ~ '^[A-Za-z0-9]+$' THEN 5
+                  WHEN s ~ '^[A-Za-z0-9._-]+$' THEN 6
+                  ELSE 7 END - 1) * 2
+          + (CASE WHEN partial THEN 1 ELSE 0 END) + 1)::smallint
+        FROM (SELECT regexp_replace(name, '\\.[A-Za-z0-9]{1,4}$', '') AS s) x
+      $$`)
+    this.schema.raw(
+      `ALTER TABLE pending_releases ADD COLUMN IF NOT EXISTS klass smallint NOT NULL DEFAULT 0`
+    )
+    this.schema.raw(`ALTER TABLE pending_releases ADD COLUMN IF NOT EXISTS target_bytes bigint`)
+    this.schema.raw(`
+      CREATE TABLE IF NOT EXISTS pending_classes (
+        tool varchar(16) NOT NULL,
+        group_id integer NOT NULL,
+        klass smallint NOT NULL,
+        probed integer NOT NULL DEFAULT 0,
+        named integer NOT NULL DEFAULT 0,
+        waiting integer NOT NULL DEFAULT 0,
+        dropped boolean NOT NULL DEFAULT false,
+        updated_at timestamptz NOT NULL DEFAULT now(),
+        PRIMARY KEY (tool, group_id, klass)
+      )`)
+    // The claim by class: the cheapest target first within one, the oldest among equals.
+    this.schema.raw(
+      `CREATE INDEX CONCURRENTLY IF NOT EXISTS pending_releases_class_idx ON pending_releases (tool, group_id, klass, target_bytes, created_at) WHERE status = 0`
+    )
+  }
+
+  async down() {
+    this.schema.raw(`DROP INDEX CONCURRENTLY IF EXISTS pending_releases_class_idx`)
+    this.schema.raw(`DROP TABLE IF EXISTS pending_classes`)
+    this.schema.raw(`ALTER TABLE pending_releases DROP COLUMN IF EXISTS target_bytes`)
+    this.schema.raw(`ALTER TABLE pending_releases DROP COLUMN IF EXISTS klass`)
+    this.schema.raw(`DROP FUNCTION IF EXISTS pending_class(text, bigint, boolean)`)
+  }
+}

database/migrations/1770000006900_pending_taken_index.ts

@@ -0,0 +1,20 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * The prober's start puts a previous run's claims back (PendingLadder.release: status 9 → 0), which scanned the
+ * 58M-row pending table on the HDD for 20+ minutes after every deploy (LOG.md (130)): a partial index on the few
+ * taken rows makes it instant.
+ */
+export default class extends BaseSchema {
+  static disableTransactions = true
+
+  async up() {
+    this.schema.raw(
+      `CREATE INDEX CONCURRENTLY IF NOT EXISTS pending_releases_taken_idx ON pending_releases (release_key) WHERE status = 9`
+    )
+  }
+
+  async down() {
+    this.schema.raw(`DROP INDEX CONCURRENTLY IF EXISTS pending_releases_taken_idx`)
+  }
+}

database/migrations/1770000007000_probe_candidates_taken_index.ts

@@ -0,0 +1,20 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * The prober's start puts a previous run's claims back (probe_candidates status taken → waiting), which scanned
+ * the 7.8 GB candidates table on the HDD for 4 minutes after every deploy (LOG.md (133)): a partial index on the
+ * few taken rows makes it instant.
+ */
+export default class extends BaseSchema {
+  static disableTransactions = true
+
+  async up() {
+    this.schema.raw(
+      `CREATE INDEX CONCURRENTLY IF NOT EXISTS probe_candidates_taken_idx ON probe_candidates (id) WHERE status = 'taken'`
+    )
+  }
+
+  async down() {
+    this.schema.raw(`DROP INDEX CONCURRENTLY IF EXISTS probe_candidates_taken_idx`)
+  }
+}

database/migrations/1770000007100_pending_swept.ts

@@ -0,0 +1,26 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * The ladder's discard sweep (PendingLadder.discard) reads a dropped signature's or class's waiting rows once:
+ * `swept_at` marks it done (null again when the record drops it anew). Before, every sweep read every dropped
+ * signature's rows over again, minutes of HDD each, inside the claim (LOG.md (133)).
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.schema.alterTable('pending_signatures', (table) => {
+      table.timestamp('swept_at', { useTz: true }).nullable()
+    })
+    this.schema.alterTable('pending_classes', (table) => {
+      table.timestamp('swept_at', { useTz: true }).nullable()
+    })
+  }
+
+  async down() {
+    this.schema.alterTable('pending_signatures', (table) => {
+      table.dropColumn('swept_at')
+    })
+    this.schema.alterTable('pending_classes', (table) => {
+      table.dropColumn('swept_at')
+    })
+  }
+}

database/migrations/1770000007200_pending_signature_index.ts

@@ -0,0 +1,23 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * The discard sweep of a dropped signature (tool, group, poster) read the whole pending_releases table per
+ * signature: the planner took a parallel seq scan over the poster test, four minutes on the HDD that starved the
+ * lanes' claims (LOG.md (134)). The waiting rows by signature make it an index range. By the poster's md5: a
+ * poster of 2.7 kB broke the first build (a btree key is at most a third of a page), which left an invalid index
+ * behind that IF NOT EXISTS would have taken for the real one; the drop clears it.
+ */
+export default class extends BaseSchema {
+  static disableTransactions = true
+
+  async up() {
+    this.schema.raw(`DROP INDEX CONCURRENTLY IF EXISTS pending_releases_signature_idx`)
+    this.schema.raw(
+      `CREATE INDEX CONCURRENTLY IF NOT EXISTS pending_releases_signature_idx ON pending_releases (group_id, tool, md5(poster)) WHERE status = 0`
+    )
+  }
+
+  async down() {
+    this.schema.raw(`DROP INDEX CONCURRENTLY IF EXISTS pending_releases_signature_idx`)
+  }
+}

database/migrations/1770000007300_probe_candidates_on_nvme.ts

@@ -0,0 +1,34 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * probe_candidates, with its indexes, in the catalog's tablespace (default_tablespace: nvme on homelab), out of
+ * CHUNK_TABLESPACE where migration 6000 put it: the prober's catalog claim updates it from random points every
+ * minute, and on the HDD, beside the pending ladder's sweep and autovacuum, a claim took 2–3 minutes with all the
+ * lanes waiting (LOG.md 2026-11-03 (136)). Homelab moved it by hand (edge, node-api and the workers stopped; ~8 GB);
+ * a no-op where it's there already, or without a default tablespace (tests).
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.defer(async (db) => {
+      const { rows: setting } = await db.rawQuery(
+        `SELECT current_setting('default_tablespace') AS target`
+      )
+      const target = setting[0].target
+      if (!target) return
+      const { rows } = await db.rawQuery(
+        `SELECT c.relname, c.relkind FROM pg_class c LEFT JOIN pg_tablespace t ON t.oid = c.reltablespace
+         WHERE (c.relname = 'probe_candidates' OR c.oid IN (SELECT indexrelid FROM pg_index WHERE indrelid = 'probe_candidates'::regclass))
+           AND c.relkind IN ('r', 'i') AND coalesce(t.spcname, 'pg_default') <> ?
+         ORDER BY c.relkind DESC, c.relname`,
+        [target]
+      )
+      for (const { relname, relkind } of rows) {
+        await db.rawQuery(
+          `ALTER ${relkind === 'i' ? 'INDEX' : 'TABLE'} ${relname} SET TABLESPACE ${target}`
+        )
+      }
+    })
+  }
+
+  async down() {}
+}

database/migrations/1770000007400_prober_tables_on_nvme.ts

@@ -0,0 +1,37 @@
+import { BaseSchema } from '@adonisjs/lucid/schema'
+
+/**
+ * Everything the prober reads or writes in the catalog's tablespace (default_tablespace: nvme on homelab), out of
+ * CHUNK_TABLESPACE (the HDD): pending_releases with its two HDD indexes (migration 6300), release_inspections
+ * (6000), probe_cache with its toast. The edge's batch upserts into pending_releases took minutes each on the HDD and
+ * the prober's flushes waited on them (LOG.md 2026-11-03 (137)); and with these on the NVMe the prober can run
+ * through the night without the HDD. Homelab moved them by hand (edge, node-api and the workers stopped; ~66 GB);
+ * a no-op where they're there already, or without a default tablespace (tests).
+ */
+export default class extends BaseSchema {
+  async up() {
+    this.defer(async (db) => {
+      const { rows: setting } = await db.rawQuery(
+        `SELECT current_setting('default_tablespace') AS target`
+      )
+      const target = setting[0].target
+      if (!target) return
+      for (const table of ['pending_releases', 'release_inspections', 'probe_cache']) {
+        const { rows } = await db.rawQuery(
+          `SELECT c.relname, c.relkind FROM pg_class c LEFT JOIN pg_tablespace t ON t.oid = c.reltablespace
+           WHERE (c.relname = ? OR c.oid IN (SELECT indexrelid FROM pg_index WHERE indrelid = ?::regclass))
+             AND c.relkind IN ('r', 'i') AND coalesce(t.spcname, 'pg_default') <> ?
+           ORDER BY c.relkind DESC, c.relname`,
+          [table, table, target]
+        )
+        for (const { relname, relkind } of rows) {
+          await db.rawQuery(
+            `ALTER ${relkind === 'i' ? 'INDEX' : 'TABLE'} ${relname} SET TABLESPACE ${target}`
+          )
+        }
+      }
+    })
+  }
+
+  async down() {}
+}

deploy/compose.yaml

@@ -126,7 +126,9 @@ services:
   prober:
     <<: *app
     command: ['node', 'ace', 'probe:run']
-    volumes: []
+    # The pending namer reads a named release's articles again from the corpus (no spool).
+    volumes:
+      - ./corpus:/corpus:ro
     environment:
       <<: *app-env
       PROBE_CONNECTIONS: ${PROBE_CONNECTIONS:-8}
@@ -173,6 +175,11 @@ services:
     environment:
       COORDINATOR_URL: http://node-api:3335
       NODE_TOKEN: ${NODE_TOKEN}
+      # The window worker (REBUILD.md design 1 and 5): a group per thread, the window's snapshot to the coordinator
+      # this often; the coordinator keeps them under SNAPSHOT_DIR (the spool's snapshots/ on bulk by default).
+      EDGE_SNAPSHOT_MINUTES: ${EDGE_SNAPSHOT_MINUTES:-30}
+      # Of the threads, this many on the groups' tail lane (the live tail; the rest work the history lane).
+      EDGE_TAIL_THREADS: ${EDGE_TAIL_THREADS:-1}
     stop_grace_period: 60s
     depends_on:
       node-api:
@@ -190,6 +197,21 @@ services:
       node-api:
         condition: service_started
 
+  # The replay harness into the scratch catalog usenet_replay on bulk (REBUILD.md, "The replay harness"):
+  # docker compose run --rm --no-deps replay --group alt.binaries.ath
+  replay:
+    <<: *app
+    profiles: [replay]
+    restart: 'no'
+    entrypoint: ['node', 'ace', 'replay:corpus']
+    command: ['--help']
+    environment:
+      <<: *app-env
+      DB_DATABASE: usenet_replay
+    volumes:
+      - ./corpus:/corpus:ro
+    depends_on: []
+
   # One-off ace commands: docker compose run --rm ace releases:duplicates
   ace:
     <<: *app

deploy/deploy.sh

@@ -47,11 +47,19 @@ echo "== recreating app services"
 docker compose stop edge
 if [ -e "$CACHE/paused" ]; then
   # Paused for the night (~/bin/usenet-rebuild-pause.sh): the workers stay stopped, the resume starts the new build.
-  echo "== paused: coordinator, prober, merger, judge, spotter, spot-inspector and edge left stopped"
-  docker compose up -d --no-deps node-api web
+  # The prober runs through the night (2026-11-03, LOG.md (137)), so it gets the new build now.
+  echo "== paused: coordinator, merger, judge, spotter, spot-inspector and edge left stopped"
+  docker compose up -d --no-deps node-api web prober
 else
   docker compose up -d --no-deps node-api web coordinator prober merger judge spotter spot-inspector
-  docker compose up -d --no-deps edge
+  if [ -e "$CACHE/no-edge" ]; then
+    # Voyager 3 (2026-10-26, LOG.md (116)): the worker indexes range by range, without the carried window, and
+    # would split releases at every range edge with nothing to join them; history comes from the corpus through
+    # deploy/replay-window.sh until the window worker is built. Remove the marker to start it again.
+    echo "== no-edge: the edge worker left stopped"
+  else
+    docker compose up -d --no-deps edge
+  fi
 fi
 echo "$SHA $(date -Is)" >> "$ROOT/deployed.log"
 # The images this replaced (321 MB each; the root filesystem is the NVMe, with a 10 GB floor).

deploy/replay-window.sh

@@ -0,0 +1,74 @@
+#!/bin/bash
+# A corpus window replayed on homelab (REBUILD.md, "The replay harness"): every group whose window ranges (the
+# ranges marked `capture` at PRIORITY in the catalog, 0 for 2016) are all stored in the corpus is replayed once
+# into the scratch catalog usenet_replay through the voyager-3 app image, PARALLEL groups at a time, the
+# window's bounds passed as --from/--to (the corpus holds other windows of the same group). The group's log
+# ends in .cache/replay/<group>.log, a .done marker next to it (.failed when the replay failed: remove it to
+# retry). Loops until every group of the window is done (the corpus fill runs meanwhile), so run it detached
+# from the deployment root:
+#   setsid nohup deploy/replay-window.sh > .cache/replay/driver.log 2>&1 &
+# No new group starts while the nightly pause is on (.cache/paused); the running ones finish.
+# DB=usenet_rebuild replays into the catalog itself (voyager 3 deployed, 2026-10-26: history comes from the corpus
+# this way until the window worker is built); the harness then needs REPLAY_INTO_CATALOG=1, passed here. OUT keeps
+# a run's logs and markers apart (.cache/replay-catalog for the catalog's).
+set -u
+ROOT=${ROOT:-/mnt/bulk/usenet-rebuild}
+PRIORITY=${PRIORITY:-0}
+PARALLEL=${PARALLEL:-4}
+MEMORY=${MEMORY:-12g}
+IMAGE=${IMAGE:-usenet-voyager3:latest}
+DB=${DB:-usenet_replay}
+OUT=${OUT:-$ROOT/.cache/replay}
+mkdir -p "$OUT"
+cd "$ROOT" || exit 1
+
+say() { echo "$(date -Is) $*"; }
+catalog() { docker compose exec -T postgres psql -U rebuild -d usenet_rebuild -X -q -At -F ' ' -c "$1"; }
+
+# The window's groups: name, first article, last article, whether every range is stored; smallest first.
+window() {
+  catalog "SELECT g.name, min(r.start_id), max(r.end_id), (count(*) = count(r.stored_at))::int
+           FROM ranges r JOIN groups g ON g.id = r.group_id
+           WHERE r.capture AND r.priority = $PRIORITY
+           GROUP BY g.name ORDER BY sum(r.end_id - r.start_id + 1)"
+}
+
+replay() {
+  local group=$1 from=$2 to=$3
+  say "$group: replaying $from–$to"
+  if docker run --rm --name "replay-${group//./-}" --network usenet-rebuild --env-file .env \
+      -e NODE_ENV=production -e HOST=0.0.0.0 -e PORT=3334 -e APP_URL=http://web:3334 -e SPOOL_DIR=/spool -e CORPUS_DIR=/corpus \
+      -e DB_HOST=postgres -e DB_PORT=5432 -e DB_DATABASE="$DB" -e CHUNK_TABLESPACE=pg_default -e REPLAY_INTO_CATALOG=1 \
+      -v "$ROOT/corpus:/corpus:ro" --memory "$MEMORY" "$IMAGE" \
+      node ace replay:corpus --group "$group" --from "$from" --to "$to" > "$OUT/$group.log" 2>&1; then
+    touch "$OUT/$group.done"
+    say "$group: done ($(grep -o 'Releases [0-9,]*' "$OUT/$group.log" | tail -1))"
+  else
+    touch "$OUT/$group.failed"
+    say "$group: FAILED (see $OUT/$group.log)"
+  fi
+  rm -f "$OUT/$group.running"
+}
+
+while true; do
+  todo=0
+  while read -r group from to complete; do
+    [ -n "$group" ] || continue
+    [ -e "$OUT/$group.done" ] && continue
+    [ -e "$OUT/$group.failed" ] && continue
+    todo=$((todo + 1))
+    [ "$complete" = 1 ] || continue
+    [ -e "$OUT/$group.running" ] && continue
+    [ -e "$ROOT/.cache/paused" ] && continue
+    while [ "$(find "$OUT" -name '*.running' | wc -l)" -ge "$PARALLEL" ]; do sleep 30; done
+    touch "$OUT/$group.running"
+    replay "$group" "$from" "$to" &
+    sleep 5
+  done <<< "$(window)"
+  if [ "$todo" = 0 ]; then
+    wait
+    say "every group of the window at priority $PRIORITY is replayed"
+    exit 0
+  fi
+  sleep 300
+done

deploy/usenet-rebuild-pause.sh

@@ -1,12 +1,13 @@
 #!/bin/sh
 # Installed as ~/bin/usenet-rebuild-pause.sh (by hand). Nightly (homelab crontab, 23:45): voyager 2's processing paused for the night; usenet-rebuild-resume.sh (07:45)
-# starts it again. Postgres, web and node-api stay up. Takes the deploy lock, so it never runs during a CI deploy, and
-# leaves .cache/paused, so a deploy in the night doesn't start the workers again.
+# starts it again. Postgres, web, node-api and the prober stay up: since 2026-11-03 (LOG.md (137)) everything the
+# prober touches is on the NVMe, so it probes through the night without the HDD. Takes the deploy lock, so it never
+# runs during a CI deploy, and leaves .cache/paused, so a deploy in the night doesn't start the workers again.
 cd /mnt/bulk/usenet-rebuild || exit 1
 exec 9>.cache/deploy.lock
 flock 9
 touch .cache/paused
 date "+%F %T pausing"
 docker compose stop edge
-docker compose stop coordinator prober merger judge spotter spot-inspector
+docker compose stop coordinator merger judge spotter spot-inspector
 docker compose ps --format "{{.Service}} {{.Status}}"

deploy/usenet-rebuild-resume.sh

@@ -7,5 +7,10 @@ flock 9
 rm -f .cache/paused
 date "+%F %T resuming"
 docker compose up -d --no-deps coordinator prober merger judge spotter spot-inspector
-docker compose up -d --no-deps edge
+if [ -e .cache/no-edge ]; then
+  # Voyager 3 without its window worker yet (deploy.sh, LOG.md (116)): the edge stays stopped.
+  echo "no-edge: the edge worker left stopped"
+else
+  docker compose up -d --no-deps edge
+fi
 docker compose ps --format "{{.Service}} {{.Status}}"

edge/Cargo.lock

@@ -44,6 +44,15 @@ version = "0.23.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ac07cdecf99051d9a5238b80f35af32cdeba5b336e55d957b318b50137e18da5"
 
+[[package]]
+name = "bincode"
+version = "1.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b1f45e9417d87227c7a56d22e471c6206462cba514c7590c09aff4cf6d1ddcad"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "block-buffer"
 version = "0.10.4"
@@ -145,6 +154,7 @@ name = "edge"
 version = "0.1.0"
 dependencies = [
  "base64 0.22.1",
+ "bincode",
  "chrono",
  "flate2",
  "libdeflater",

edge/Cargo.toml

@@ -2,7 +2,7 @@
 name = "edge"
 version = "0.1.0"
 edition = "2021"
-description = "Voyager 2's edge worker (edge-node) and its benchmark (edge-bench): the RangeIndexer in Rust, kept identical to lib/edge (check.ts)"
+description = "The edge worker (edge-node) and its benchmark (edge-bench): voyager 3's window indexing in Rust"
 
 [[bin]]
 name = "edge-bench"
@@ -14,6 +14,7 @@ path = "src/bin/edge-node.rs"
 
 [dependencies]
 base64 = "0.22"
+bincode = "1.3"
 chrono = "0.4"
 flate2 = { version = "1", default-features = false, features = ["zlib-rs"] }
 libdeflater = "1.26.1"

edge/README.md

@@ -6,18 +6,21 @@ from the node API, reads their prefetched XOVER chunks, finds releases and
 posts batches. `edge-bench` (`src/main.rs`) replays a captured range through
 the same indexing (`src/lib.rs`) for timing and for comparing.
 
-The TypeScript `RangeIndexer` (`lib/edge`, `node ace edge:node`) is the
-reference: `check.ts` runs both on a capture and compares every output field
-by field. An identity or rule change goes into both and has to come out
-identical. CI (`.gitea/workflows/deploy.yaml`, job `edge`) runs `cargo test`
-and `check-fixtures.sh`, which compares both on the committed captures in
-`fixtures/` (25k overview lines each of a.b.models, a.b.ath and
-a.b.boneless, in 5 chunks); a push to `rebuild` deploys only when they
-match.
+Since voyager 3 (2026-10-25) the Rust indexer is the one implementation of
+the window (REBUILD.md, "Voyager 3: window indexing"; `src/indexer.rs`):
+files and releases are finished when the frontier (the median Date of the
+last 1,024 plausibly dated articles) is 24 h past their last part, or at
+the block's end (provisionally, as edges), and written once. The
+TypeScript `RangeIndexer` (`lib/edge`) keeps the range path for the corpus
+scanner and the tests; the field-by-field comparison (`check.ts`) went
+with the parity. CI (`.gitea/workflows/deploy.yaml`, job `edge`) runs
+`cargo test` and `bench-fixtures.sh`, which replays the committed captures
+in `fixtures/` (25k overview lines each of a.b.ath, a.b.boneless, a.b.fta
+and a.b.models, in 5 chunks).
 
 ```sh
 cargo test --release --manifest-path edge/Cargo.toml
-edge/check-fixtures.sh                       # from the repository root
+edge/bench-fixtures.sh                       # from the repository root
 ```
 
 In production since 2026-10-09: 122M headers in 10 minutes with 0.37 cores
@@ -110,11 +113,30 @@ files, segments, payload size, candidates, fragments and samples. The
 summary can hide compensating differences, so compare the outputs
 themselves.
 
-## Comparing the output
+## Replaying a group through one window
+
+A corpus group directory (`<corpus>/<group>/<start>-<end>/`, each range with
+its `range.json`) is replayed in article order with the window carried from
+range to range, the end finished after the last range (REBUILD.md, "The
+replay harness"):
+
+```sh
+./target/release/edge-bench --group /corpus/alt.binaries.ath --rules rules.json --batch-chunks 5 --wire - > batches.jsonl
+# --from <article> skips ranges ending before it, --to <article> ranges starting after it (one window of the
+# group's history), --limit <n> stops after n ranges, --no-finish leaves the window open
+```
+
+`--rules` takes the lease's rules JSON (`src/lease.rs`: the catalog's rule
+set and group priorities, as `replay:corpus` writes it); without it the
+defaults with only the group indexed apply. `--wire -` writes each batch as
+the worker posts it to stdout, one JSON per line, and the report to stderr;
+`node ace replay:corpus` reads them into the batch store. Each range end
+prints the frontier, the releases out and the releases and files still open.
+
+## The dump
 
 ```sh
 ./target/release/edge-bench /path/to/capture --dump /tmp/out.jsonl
-cd .. && node --import=@poppinss/ts-exec edge/check.ts /path/to/capture /tmp/out.jsonl
 ```
 
 `--dump` writes every output of every batch as JSON lines:
@@ -212,17 +234,45 @@ them:
 
 ```sh
 cargo build --release
-COORDINATOR_URL=http://coordinator:3334 NODE_TOKEN=… ./target/release/edge-node [--threads N] [--batch-articles 500000] [--ranges N] [--max-mbit M]
+COORDINATOR_URL=http://coordinator:3334 NODE_TOKEN=… ./target/release/edge-node [--threads N] [--tail-threads T] [--batch-articles 500000] [--snapshot-minutes 30] [--groups N] [--max-mbit M]
 ```
 
-- **One range per thread** (default one per core): one thread reads the
-  coordinator's stream and inflates a frame ahead, another posts the
-  batches in order (one may wait), the indexer goes on meanwhile.
+- **Two lanes** (2026-10-27): `--tail-threads T` (EDGE_TAIL_THREADS; one of
+  two or more threads by default) puts that many threads on the groups'
+  *tail* lane (`?lane=tail` on the group routes): the tail planner's
+  ranges, through a window and snapshot of their own
+  (`<group>-tail.gz`). The other threads work the history lane: the
+  windows and the history the coordinator's bounds hand out. A tail thread
+  never wades through a group's history first, and a held history group's
+  tail isn't stuck behind its days of ranges; the coordinator orders each
+  lane's leases by alternating the group with the most planned articles
+  and the longest unheld one.
+
+- **One group per thread** (default one per core; REBUILD.md, "Voyager 3:
+  window indexing", 1 and 5): a thread leases a group (`POST
+  groups/lease`), takes its ranges in article order (`POST
+  groups/:id/next`), streams each range's chunks (`GET ranges/:id/stream`:
+  the corpus or the spool, the worker can't tell) through one carried
+  window, posts a batch at least at every range end and finishes the
+  range; one thread reads the stream and inflates a frame ahead, another
+  posts the batches in order, the indexer goes on meanwhile. The window's
+  snapshot (bincode, gzip) goes to the coordinator every
+  `--snapshot-minutes` (EDGE_SNAPSHOT_MINUTES, 30) right after a batch,
+  and when the group is given back: caught up (`next` answers 204: the
+  tail), a stop, a lost lease or a failure; a stop or a failed stream
+  posts the open batch first, so the snapshot stands at its resume point
+  (a batch the coordinator refused can't, and that window is lost); the
+  next holder restores it (`GET groups/:id/snapshot`) and continues at
+  its article. `--groups N`
+  stops after N groups; `--ranges N` is the old per-range lease
+  (`POST lease`), one range per thread without a carried window.
 - **As `commands/edge_node.ts`:** retries for network trouble and 5xx;
   409 means the lease is lost (the range stops, isn't given back); a
   batch once 500k articles or 100k files wait (a picture group's batches
   went over the server's 256 MB limit); a chunk that doesn't inflate is
-  reported (the coordinator fetches it plain next time); SIGTERM/SIGINT
+  reported (the coordinator fetches it plain next time) after the batch
+  before it went out, and a held group asks for the range again from
+  that chunk, its window kept; SIGTERM/SIGINT
   give each range back after its current chunk; `--max-mbit` is the
   node's download cap, shared by its threads. Plus a heartbeat when no
   batch went out for a minute.
@@ -234,5 +284,11 @@ COORDINATOR_URL=http://coordinator:3334 NODE_TOKEN=… ./target/release/edge-nod
   same (releases with every column, chunks, candidates, rule counts,
   fragments with their JSON).
 - **On homelab:** compose service `edge` (deploy/compose.yaml), the
-  `edge` image target of the Dockerfile, `--threads $EDGE_THREADS` (8);
-  it talks to `node-api`.
+  `edge` image target of the Dockerfile, `--threads $EDGE_THREADS` (8),
+  EDGE_TAIL_THREADS (1); it talks to `node-api`.
+- **End to end:** `tests/e2e/edge_node.spec.ts` runs this binary
+  (`edge/target/release/edge-node`, or EDGE_NODE_BIN) against the test
+  server, database, spool and corpus: a chunk the corpus fill stores while
+  the stream waits, a failed stream and the restart that restores the
+  snapshot, a restart whose snapshot file is gone, a takeover. `node ace
+  test e2e` after `cargo build --release`; CI builds the binary first.

edge/bench-fixtures.sh

@@ -0,0 +1,11 @@
+#!/bin/sh
+# The Rust indexer over the committed captures (edge/fixtures: 25k overview lines each of a.b.ath, a.b.boneless,
+# a.b.fta and a.b.models, in 5 chunks): a smoke test of the window on real headers, exiting non-zero when the bench
+# fails on one. CI runs it after cargo test; run it from the repository root after a change to the indexer.
+set -eu
+cd "$(dirname "$0")/.."
+cargo build --release --quiet --manifest-path edge/Cargo.toml --bin edge-bench
+for capture in edge/fixtures/*/; do
+  echo "== $(basename "$capture")"
+  edge/target/release/edge-bench "$capture" --batch-chunks 2 --idle 2 | tail -1
+done

edge/check-fixtures.sh

@@ -1,18 +0,0 @@
-#!/bin/sh
-# The Rust and TypeScript RangeIndexers on the committed captures (edge/fixtures: 25k
-# overview lines each of a.b.models, a.b.ath and a.b.boneless, in 5 chunks), compared
-# by check.ts field by field: exits non-zero on any difference. CI runs it; run it from
-# the repository root after any change to either side.
-set -eu
-cd "$(dirname "$0")/.."
-cargo build --release --quiet --manifest-path edge/Cargo.toml --bin edge-bench
-out=$(mktemp -d)
-trap 'rm -rf "$out"' EXIT
-status=0
-for capture in edge/fixtures/*/; do
-  name=$(basename "$capture")
-  edge/target/release/edge-bench "$capture" --batch-chunks 2 --idle 2 --dump "$out/$name.jsonl" --wire "$out/$name.wire.jsonl" >/dev/null 2>&1
-  echo "== $name"
-  node --import=@poppinss/ts-exec edge/check.ts "$capture" "$out/$name.jsonl" --wire "$out/$name.wire.jsonl" --batch-chunks 2 --idle 2 || status=1
-done
-exit $status

edge/check.ts

@@ -1,139 +0,0 @@
-/**
- * Compares edge-bench's output (--dump) with the TypeScript RangeIndexer
- * on the same captured range, item by item: releases (every field, and the
- * payload decoded with lib/chunks), samples, probe candidates, rule counts,
- * fragments and batch boundaries.
- *
- *   node --import=@poppinss/ts-exec edge/check.ts <dir> <dump.jsonl> [--wire batches.jsonl] [--batch-chunks 10] [--idle 5]
- *
- * With --wire (edge-bench --wire): also each batch as the Rust worker
- * posts it, against toWire of the TypeScript batch, field by field.
- */
-import { readFile, readdir } from 'node:fs/promises'
-import { join } from 'node:path'
-import { unzipSync } from 'node:zlib'
-import { isDeepStrictEqual } from 'node:util'
-import { RangeIndexer, toWire, type EdgeBatch } from '#lib/edge/index'
-import { DEFAULT_RULES, type Rules } from '#lib/rules/index'
-import { decodeChunk } from '#lib/chunks/index'
-
-const [dir, dumpPath] = process.argv.slice(2)
-const flag = (name: string, fallback: number) => {
-  const i = process.argv.indexOf(name)
-  return i === -1 ? fallback : Number(process.argv[i + 1])
-}
-const batchChunks = flag('--batch-chunks', 10)
-const idle = flag('--idle', 5)
-
-const { group, start, end } = JSON.parse(await readFile(join(dir, 'range.json'), 'utf8'))
-const rules: Rules = { ...DEFAULT_RULES, priorities: { [group]: 1 }, droppedSignatures: [] }
-const indexer = new RangeIndexer({ group, start, end, rules, idleChunks: idle })
-const files = (await readdir(dir)).filter((f) => f.endsWith('.gz')).sort()
-const batches: EdgeBatch[] = []
-for (const [i, file] of files.entries()) {
-  const next = Number(file.split('-')[2].replace('.gz', '')) + 1
-  indexer.addText(unzipSync(await readFile(join(dir, file))).toString('utf8'), next)
-  const last = i === files.length - 1
-  if (last) indexer.finish()
-  if (last || (i + 1) % batchChunks === 0) batches.push(indexer.take())
-}
-
-/** Canonical lines per kind, in a stable order, from both sides. */
-type Item = Record<string, unknown>
-const sortFiles = (decoded: ReturnType<typeof decodeChunk>) => [...decoded].sort((a, b) => (a.name + a.subject < b.name + b.subject ? -1 : 1))
-const payloadBytes = { ts: 0, rs: 0 }
-let side: 'ts' | 'rs' = 'ts'
-const release = (r: Item, b: number) => {
-  const { payload, ...rest } = r as Item & { payload: Buffer }
-  // The compressed size isn't compared (node's zstd and the zstd crate frame alike data a little
-  // differently); what the payload decodes to is.
-  payloadBytes[side] += payload.length
-  return { b, ...rest, decoded: sortFiles(decodeChunk(payload)) }
-}
-const ts: Record<string, Item[]> = { b: [], r: [], s: [], c: [], n: [], f: [] }
-batches.forEach((batch, b) => {
-  ts.b.push({ b, from: batch.from, next: batch.next, headers: batch.headers, postedLow: batch.postedLow, postedHigh: batch.postedHigh })
-  for (const r of batch.releases) ts.r.push(release({ upload: r.upload, key: r.key, name: r.name, source: r.source, fileTotal: r.fileTotal, poster: r.poster, groups: r.groups, postedAt: r.postedAt, postedLast: r.postedLast, files: r.files, segments: r.segments, bytes: r.bytes, firstArticle: r.firstArticle, lastArticle: r.lastArticle, readable: r.readable, hasPar2: r.hasPar2, payload: r.payload }, b))
-  for (const s of batch.samples) ts.s.push({ b, key: s.key, upload: s.upload, rule: s.rule, ruleKey: s.ruleKey, segments: s.segments })
-  for (const c of batch.candidates) ts.c.push({ b, upload: c.upload, releaseKey: c.releaseKey, kind: c.kind, messageId: c.messageId, file: c.file })
-  for (const n of batch.counts) ts.n.push({ b, rule: n.rule, key: n.key, articles: n.articles, files: n.files })
-  for (const f of batch.fragments) ts.f.push({ b, fileKey: f.fileKey, name: f.name, totalParts: f.totalParts, postedAt: f.postedAt, segments: f.segments })
-})
-
-const rs: Record<string, Item[]> = { b: [], r: [], s: [], c: [], n: [], f: [] }
-side = 'rs'
-let b = -1
-for (const line of (await readFile(dumpPath, 'utf8')).split('\n')) {
-  if (!line) continue
-  const { t, ...item } = JSON.parse(line)
-  if (t === 'b') b++
-  if (t === 'r') rs.r.push(release({ ...item, payload: Buffer.from(item.payload, 'base64') }, b))
-  else if (t === 's') rs.s.push({ b, key: item.key, upload: item.upload, rule: item.rule, ruleKey: item.ruleKey, segments: item.segments })
-  else rs[t].push({ b, ...item })
-}
-
-const keyOf: Record<string, (i: Item) => string> = {
-  b: (i) => String(i.b),
-  r: (i) => `${i.b} ${i.key} ${i.upload}`,
-  s: (i) => `${i.b} ${i.key} ${i.upload}`,
-  c: (i) => `${i.b} ${i.releaseKey} ${i.upload} ${i.kind} ${i.messageId}`,
-  n: (i) => `${i.b} ${i.rule} ${i.key}`,
-  f: (i) => `${i.b} ${i.fileKey}`,
-}
-const names: Record<string, string> = { b: 'batches', r: 'releases', s: 'samples', c: 'candidates', n: 'rule counts', f: 'fragments' }
-let failed = false
-for (const kind of Object.keys(keyOf)) {
-  const a = new Map(ts[kind].map((i) => [keyOf[kind](i), i]))
-  const z = new Map(rs[kind].map((i) => [keyOf[kind](i), i]))
-  const missing = [...a.keys()].filter((k) => !z.has(k))
-  const extra = [...z.keys()].filter((k) => !a.has(k))
-  const differ = [...a.keys()].filter((k) => z.has(k) && !isDeepStrictEqual(a.get(k), z.get(k)))
-  const ok = missing.length === 0 && extra.length === 0 && differ.length === 0
-  failed ||= !ok
-  console.log(`${ok ? 'same' : 'DIFF'} ${names[kind]}: ${a.size} TypeScript, ${z.size} Rust${ok ? '' : `; ${missing.length} only in TypeScript, ${extra.length} only in Rust, ${differ.length} differ`}`)
-  for (const k of [...missing.slice(0, 2), ...extra.slice(0, 2)]) console.log(`  only one side: ${k}`)
-  for (const k of differ.slice(0, 2)) {
-    const x = a.get(k)!
-    const y = z.get(k)!
-    const fields = Object.keys(x).filter((f) => !isDeepStrictEqual(x[f], y[f]))
-    console.log(`  ${k}: ${fields.map((f) => `${f}: ${JSON.stringify(x[f]).slice(0, 160)} vs ${JSON.stringify(y[f]).slice(0, 160)}`).join('; ')}`)
-  }
-}
-console.log(`payload bytes: ${payloadBytes.ts} TypeScript, ${payloadBytes.rs} Rust`)
-
-const wireIndex = process.argv.indexOf('--wire')
-if (wireIndex !== -1) {
-  // Order within a batch doesn't count; payloads by what they decode to.
-  const by = (f: (x: any) => string) => (a: any, b: any) => (f(a) < f(b) ? -1 : f(a) > f(b) ? 1 : 0)
-  const normal = (w: any) => ({
-    ...w,
-    releases: w.releases.map((r: any) => ({ ...r, payload: sortFiles(decodeChunk(Buffer.from(r.payload, 'base64'))) })).sort(by((r) => `${r.key} ${r.upload}`)),
-    samples: w.samples.map((r: any) => ({ ...r, payload: sortFiles(decodeChunk(Buffer.from(r.payload, 'base64'))) })).sort(by((r) => `${r.key} ${r.upload}`)),
-    candidates: [...w.candidates].sort(by((c) => JSON.stringify(c))),
-    counts: [...w.counts].sort(by((c) => `${c.rule} ${c.key}`)),
-    fragments: [...w.fragments].sort(by((f) => `${f.fileKey} ${f.segments[0]?.[1]}`)),
-  })
-  const rust = (await readFile(process.argv[wireIndex + 1], 'utf8')).split('\n').filter(Boolean).map((l) => normal(JSON.parse(l)))
-  const typescript = batches.map((b) => normal(JSON.parse(JSON.stringify(toWire(b)))))
-  let wireSame = rust.length === typescript.length
-  const firstDiff = (x: any, y: any, path: string): string | null => {
-    if (isDeepStrictEqual(x, y)) return null
-    if (typeof x !== 'object' || typeof y !== 'object' || x === null || y === null) return `${path}: ${JSON.stringify(x)?.slice(0, 120)} vs ${JSON.stringify(y)?.slice(0, 120)}`
-    for (const k of new Set([...Object.keys(x), ...Object.keys(y)])) {
-      const d = firstDiff(x[k], y[k], `${path}.${k}`)
-      if (d) return d
-    }
-    return path
-  }
-  for (let i = 0; i < Math.max(rust.length, typescript.length); i++) {
-    const d = firstDiff(typescript[i], rust[i], `batch ${i}`)
-    if (d) {
-      wireSame = false
-      console.log(`DIFF wire: ${d}`)
-      break
-    }
-  }
-  if (wireSame) console.log(`same wire batches: ${typescript.length} TypeScript, ${rust.length} Rust`)
-  failed ||= !wireSame
-}
-process.exit(failed ? 1 : 0)

edge/src/bin/edge-node.rs

@@ -1,16 +1,24 @@
-//! The edge worker in Rust (REBUILD.md, part 2; commands/edge_node.ts): leases
-//! ranges from the coordinator, reads their XOVER chunks from its stream,
-//! indexes them (the same RangeIndexer as edge-bench, kept identical to
-//! the TypeScript by check.ts) and posts the batches. One range per thread.
+//! The window worker in Rust (REBUILD.md, "Voyager 3: window indexing", 1 and 5): each thread leases a *group*
+//! from the coordinator, takes its ranges in article order, reads each range's XOVER chunks from the coordinator's
+//! stream (the corpus or the spool, the worker can't tell), indexes them through one carried window and posts the
+//! batches, a batch at least at every range end; the window's snapshot goes to the coordinator every
+//! --snapshot-minutes and when the group is given back, and the next holder restores it. A group with no range
+//! left (caught up: the tail) is given back and leased again later.
 //!
-//!   COORDINATOR_URL=… NODE_TOKEN=… edge-node [--threads N] [--batch-articles 500000] [--ranges N] [--max-mbit M]
+//!   COORDINATOR_URL=… NODE_TOKEN=… edge-node [--threads N] [--tail-threads T] [--batch-articles 500000] [--snapshot-minutes 30] [--groups N] [--max-mbit M]
 //!
-//! SIGTERM/SIGINT: each thread gives its range back after the current chunk
-//! (the next worker resumes at the last stored batch).
+//! `--tail-threads T` (EDGE_TAIL_THREADS; 1 of 2 or more threads by default): that many of the threads lease the
+//! groups' *tail* lane (`?lane=tail`: the tail planner's ranges, with a window and snapshot of their own), the rest
+//! the history lane (the windows and the history the coordinator's bounds hand out).
+//!
+//! SIGTERM/SIGINT: each thread finishes the chunk, posts its batch, snapshots and gives its group back.
+//! `--ranges`: the old range leases (POST lease), one range per thread without a carried window (kept for a
+//! coordinator without group leases).
 
 use edge::indexer::{Batch, RangeIndexer};
-use edge::names::Rules;
+use edge::lease::{GroupLeaseJson, LeaseJson, NextRangeJson, RangeSpanJson};
 use edge::wire;
+use flate2::read::GzDecoder;
 use flate2::write::GzEncoder;
 use serde::Deserialize;
 use std::collections::HashMap;
@@ -34,63 +42,6 @@ fn log(level: &str, message: &str) {
     println!("[ {level} ] {message}");
 }
 
-#[derive(Deserialize)]
-struct LeaseJson {
-    range: RangeJson,
-    rules: RulesJson,
-}
-
-#[derive(Deserialize)]
-struct RangeJson {
-    id: i64,
-    group: String,
-    start: i64,
-    end: i64,
-    next: i64,
-}
-
-#[derive(Deserialize)]
-#[serde(rename_all = "camelCase")]
-struct RulesJson {
-    version: i64,
-    priorities: HashMap<String, u32>,
-    #[serde(default)]
-    dropped_signatures: Vec<String>,
-    #[serde(default)]
-    dropped_tools: Option<Vec<String>>,
-    #[serde(default)]
-    per_article_max_parts: Option<u32>,
-    #[serde(default)]
-    sample_percent: Option<f64>,
-    #[serde(default)]
-    drop_executables: Option<bool>,
-    #[serde(default)]
-    dropped_posters: Option<Vec<String>>,
-}
-
-impl RulesJson {
-    fn rules(self) -> Rules {
-        let mut rules = Rules::defaults(self.priorities.into_iter().collect());
-        rules.version = self.version;
-        rules.dropped_signatures = self.dropped_signatures.into_iter().collect();
-        if let Some(tools) = self.dropped_tools {
-            rules.dropped_tools = tools.into_iter().collect();
-        }
-        if let Some(n) = self.per_article_max_parts {
-            rules.per_article_max_parts = n;
-        }
-        if let Some(p) = self.sample_percent {
-            rules.sample_percent = p;
-        }
-        if let Some(e) = self.drop_executables {
-            rules.drop_executables = e;
-        }
-        if let Some(posters) = self.dropped_posters {
-            rules.dropped_posters = posters;
-        }
-        rules
-    }
-}
 
 #[derive(Deserialize)]
 struct StoredJson {
@@ -220,6 +171,81 @@ impl Api {
         }
     }
 
+    // Group leases (REBUILD.md design 1 and 5), per lane (`history` or `tail`).
+
+    fn lease_group(&self, lane: &str) -> Result<Option<GroupLeaseJson>, ApiError> {
+        let mut response = self.request("POST", &format!("groups/lease?lane={lane}"), None, false, "application/json")?;
+        match response.status().as_u16() {
+            204 => Ok(None),
+            200 => {
+                let text = response.body_mut().read_to_string().map_err(|e| ApiError::Other(format!("groups/lease: {e}")))?;
+                serde_json::from_str(&text).map(Some).map_err(|e| ApiError::Other(format!("groups/lease: {e}")))
+            }
+            status => Err(ApiError::Other(format!("groups/lease: HTTP {status}"))),
+        }
+    }
+
+    fn next_range(&self, group: i64, lane: &str) -> Result<Option<RangeSpanJson>, ApiError> {
+        let mut response = self.request("POST", &format!("groups/{group}/next?lane={lane}"), None, false, "application/json")?;
+        match response.status().as_u16() {
+            204 => Ok(None),
+            200 => {
+                let text = response.body_mut().read_to_string().map_err(|e| ApiError::Other(format!("next: {e}")))?;
+                serde_json::from_str::<NextRangeJson>(&text).map(|n| Some(n.range)).map_err(|e| ApiError::Other(format!("next: {e}")))
+            }
+            status => Err(ApiError::Other(format!("next: HTTP {status}"))),
+        }
+    }
+
+    fn group_call(&self, group: i64, lane: &str, action: &str) -> Result<(), ApiError> {
+        let response = self.request("POST", &format!("groups/{group}/{action}?lane={lane}"), None, false, "application/json")?;
+        match response.status().as_u16() {
+            200..=299 => Ok(()),
+            status => Err(ApiError::Other(format!("{action}: HTTP {status}"))),
+        }
+    }
+
+    /// The window's snapshot (bincode), gzipped, standing at `next`.
+    fn put_snapshot(&self, group: i64, lane: &str, next: i64, snapshot: &[u8]) -> Result<usize, ApiError> {
+        let mut gz = GzEncoder::new(Vec::with_capacity(snapshot.len() / 3), flate2::Compression::new(1));
+        gz.write_all(snapshot).unwrap();
+        let body = gz.finish().unwrap();
+        let url = format!("{}/api/v1/node/groups/{group}/snapshot?next={next}&lane={lane}", self.base);
+        let auth = format!("Bearer {}", self.token);
+        for attempt in 1..=RETRIES {
+            let result = self.agent.put(&url).header("Authorization", &auth).header("Content-Type", "application/octet-stream").send(&body[..]);
+            match result {
+                Ok(response) => match response.status().as_u16() {
+                    200..=299 => return Ok(body.len()),
+                    409 => return Err(ApiError::LeaseLost("snapshot: the group is no longer this node's".into())),
+                    status if status < 500 => return Err(ApiError::Other(format!("snapshot: HTTP {status}"))),
+                    status if attempt == RETRIES => return Err(ApiError::Other(format!("snapshot: HTTP {status}"))),
+                    _ => {}
+                },
+                Err(error) if attempt == RETRIES => return Err(ApiError::Other(format!("snapshot: {error}"))),
+                Err(_) => {}
+            }
+            std::thread::sleep(Duration::from_millis(60_000.min(1000 * 2u64.pow(attempt))));
+        }
+        unreachable!()
+    }
+
+    /// The group's snapshot: its bytes (inflated) and the article it stands at; None without one.
+    fn get_snapshot(&self, group: i64, lane: &str) -> Result<Option<(i64, Vec<u8>)>, ApiError> {
+        let response = self.request("GET", &format!("groups/{group}/snapshot?lane={lane}"), None, false, "application/octet-stream")?;
+        match response.status().as_u16() {
+            404 => Ok(None),
+            200 => {
+                let next: i64 = response.headers().get("x-snapshot-next").and_then(|v| v.to_str().ok()).and_then(|v| v.parse().ok()).ok_or_else(|| ApiError::Other("snapshot: no X-Snapshot-Next".into()))?;
+                let mut gz = GzDecoder::new(response.into_body().into_reader());
+                let mut bytes = Vec::new();
+                gz.read_to_end(&mut bytes).map_err(|e| ApiError::Other(format!("snapshot: {e}")))?;
+                Ok(Some((next, bytes)))
+            }
+            status => Err(ApiError::Other(format!("snapshot: HTTP {status}"))),
+        }
+    }
+
     fn corrupt(&self, range: i64, from: i64) -> Result<(), ApiError> {
         let body = format!("{{\"from\":{from}}}");
         let response = self.request("POST", &format!("ranges/{range}/corrupt"), Some(body.as_bytes()), false, "application/json")?;
@@ -332,6 +358,8 @@ fn inflate(raw: &[u8], d: &mut libdeflater::Decompressor) -> Result<Vec<u8>, Str
 struct Config {
     batch_articles: i64,
     bytes_per_second: Option<f64>,
+    /// Between two snapshots of a held group's window.
+    snapshot_every: Duration,
     /// Ranges whose batch was too large (413), and how much smaller their batches are now (halved each time).
     smaller: Mutex<HashMap<i64, i64>>,
 }
@@ -509,6 +537,272 @@ fn run_range(api: &Arc<Api>, lease: LeaseJson, config: &Config, stopping: &Atomi
     }
 }
 
+/// What one range of a held group came to.
+enum RangeOutcome {
+    Done,
+    /// A chunk didn't inflate: reported (the coordinator fetches it again, plain the second time) after the batch
+    /// before it went out, and the range is asked for again from there; the window stays.
+    Again,
+    /// The group is given back (a stop, a lost lease, a failure): no more ranges.
+    Stop,
+}
+
+/// One range of a held group through the carried window: the chunks streamed, batches posted at the range's end at
+/// least, the range finished. The indexer keeps the window for the next range. `group_name` is the label for the
+/// log (the tail lane's says so).
+#[allow(clippy::too_many_arguments)]
+fn run_group_range(api: &Arc<Api>, group_id: i64, lane: &str, group_name: &str, range: RangeSpanJson, indexer: &mut RangeIndexer, config: &Config, stopping: &AtomicBool, totals: &Arc<Mutex<Totals>>) -> RangeOutcome {
+    let started = Instant::now();
+    indexer.begin_range(range.next, range.end);
+    let post_error: Arc<Mutex<Option<ApiError>>> = Arc::new(Mutex::new(None));
+    let last_contact = Arc::new(AtomicI64::new(0));
+    let since = Instant::now();
+    let (post_tx, post_rx) = sync_channel::<Batch>(1);
+    let poster = {
+        let (api, totals, post_error, last_contact) = (api.clone(), totals.clone(), post_error.clone(), last_contact.clone());
+        std::thread::spawn(move || {
+            for batch in post_rx {
+                if post_error.lock().unwrap().is_some() {
+                    continue;
+                }
+                let at = Instant::now();
+                match api.batch(range.id, &batch) {
+                    Ok((releases, candidates, bytes)) => {
+                        let mut t = totals.lock().unwrap();
+                        t.post_ms += at.elapsed().as_millis();
+                        t.bytes += bytes;
+                        t.headers += batch.headers;
+                        t.releases += releases;
+                        t.candidates += candidates;
+                        last_contact.store(since.elapsed().as_millis() as i64, Ordering::Relaxed);
+                    }
+                    Err(e) => *post_error.lock().unwrap() = Some(e),
+                }
+            }
+        })
+    };
+    let frames = match api.stream(range.id) {
+        Ok(stream) => read_frames(stream, config.bytes_per_second),
+        Err(e) => {
+            log("error", &format!("{group_name} {}–{}: stream failed: {e}; the group goes back.", range.start, range.end));
+            drop(post_tx);
+            let _ = poster.join();
+            return RangeOutcome::Stop;
+        }
+    };
+    let (batch_articles, max_pending) = config.limits(range.id);
+    let mut batch_from = range.next;
+    let mut reached_end = false;
+    let mut corrupt: Option<(i64, i64, String)> = None;
+    let mut failure: Option<String> = None;
+    let headers_before = totals.lock().unwrap().headers;
+    for frame in frames.iter() {
+        if stopping.load(Ordering::Relaxed) || post_error.lock().unwrap().is_some() {
+            break;
+        }
+        if since.elapsed().as_millis() as i64 - last_contact.load(Ordering::Relaxed) > HEARTBEAT.as_millis() as i64 {
+            last_contact.store(since.elapsed().as_millis() as i64, Ordering::Relaxed);
+            let _ = api.group_call(group_id, lane, "heartbeat");
+        }
+        match frame {
+            Frame::KeepAlive => continue,
+            Frame::End => break,
+            Frame::Failed(e) => {
+                failure = Some(e);
+                break;
+            }
+            Frame::Chunk { from, to, text } => {
+                let data = match text {
+                    Ok(data) => data,
+                    Err(e) => {
+                        corrupt = Some((from, to, e));
+                        break;
+                    }
+                };
+                indexer.add_text(&String::from_utf8_lossy(&data), to + 1);
+                reached_end = to >= range.end;
+                // The window carries on past the range's end: no finish here; a batch at the end at least.
+                if reached_end || to + 1 - batch_from >= batch_articles || indexer.pending() >= max_pending {
+                    if post_tx.send(indexer.take()).is_err() {
+                        break;
+                    }
+                    batch_from = to + 1;
+                }
+                if reached_end {
+                    break;
+                }
+            }
+        }
+    }
+    // Stopped mid-range (a stop, a stream that failed, a chunk that doesn't inflate): the batch so far goes out,
+    // so the snapshot stands at a stored resume point and the window survives the group going back; a chunk
+    // reported corrupt is read again from its own start. Only a batch the coordinator refused stays open.
+    if !reached_end && post_error.lock().unwrap().is_none() && indexer.next() > batch_from {
+        let _ = post_tx.send(indexer.take());
+    }
+    drop(post_tx);
+    let _ = poster.join();
+    let post_error = post_error.lock().unwrap().take();
+    if let Some((from, to, e)) = corrupt {
+        // The server's compressed reply cut short (LOG.md 2026-10-11): the coordinator fetches the chunk again,
+        // plain the second time, and sets the range aside after that; the window carries on either way.
+        log("warn", &format!("{group_name} {}–{}: chunk {from}–{to} doesn't inflate ({e}); reported, the range is asked for again", range.start, range.end));
+        if let Err(e) = api.corrupt(range.id, from) {
+            log("warn", &format!("{group_name} {}–{}: reporting the chunk: {e}; the group goes back", range.start, range.end));
+            return RangeOutcome::Stop;
+        }
+        if post_error.is_none() {
+            return RangeOutcome::Again;
+        }
+    }
+    match post_error {
+        Some(ApiError::LeaseLost(m)) => {
+            log("warn", &format!("{group_name} {}–{}: {m}", range.start, range.end));
+            return RangeOutcome::Stop;
+        }
+        Some(ApiError::TooLarge(m)) => {
+            let divisor = {
+                let mut smaller = config.smaller.lock().unwrap();
+                let d = smaller.entry(range.id).or_insert(1);
+                *d = (*d * 2).min(64);
+                *d
+            };
+            log("warn", &format!("{group_name} {}–{}: {m}; the group goes back, its batches 1/{divisor} the size from now on.", range.start, range.end));
+            return RangeOutcome::Stop;
+        }
+        Some(ApiError::Other(m)) => {
+            log("error", &format!("{group_name} {}–{}: {m}; the group goes back.", range.start, range.end));
+            return RangeOutcome::Stop;
+        }
+        None => {}
+    }
+    if let Some(e) = failure {
+        log("error", &format!("{group_name} {}–{}: {e}; the group goes back.", range.start, range.end));
+        return RangeOutcome::Stop;
+    }
+    if !reached_end {
+        return RangeOutcome::Stop;
+    }
+    match api.call(range.id, "finish") {
+        Ok(()) => {
+            let t = totals.lock().unwrap();
+            let seconds = started.elapsed().as_secs_f64();
+            let headers = t.headers - headers_before;
+            log(
+                "info",
+                &format!(
+                    "{group_name} {}–{}: {} headers in {:.0} s ({:.0}/s); {} releases and {} files open in the window.",
+                    range.start,
+                    range.end,
+                    headers,
+                    seconds,
+                    headers as f64 / seconds.max(0.001),
+                    indexer.open_releases(),
+                    indexer.open_files()
+                ),
+            );
+            RangeOutcome::Done
+        }
+        Err(e) => {
+            log("warn", &format!("{group_name} {}–{}: finish: {e}", range.start, range.end));
+            RangeOutcome::Stop
+        }
+    }
+}
+
+/// The window's snapshot to the coordinator, standing at the indexer's next article.
+fn snapshot(api: &Api, group_id: i64, lane: &str, group_name: &str, indexer: &RangeIndexer) {
+    match indexer.snapshot() {
+        Ok(bytes) => match api.put_snapshot(group_id, lane, indexer.next(), &bytes) {
+            Ok(gz) => log("info", &format!("{group_name}: snapshot at {} ({:.1} MB, {:.1} MB gzipped).", indexer.next(), bytes.len() as f64 / 1e6, gz as f64 / 1e6)),
+            Err(e) => log("warn", &format!("{group_name}: snapshot failed: {e}")),
+        },
+        Err(e) => log("warn", &format!("{group_name}: no snapshot: {e}")),
+    }
+}
+
+/// One held group (one lane of it): its window restored from the snapshot (or new), its ranges one after another
+/// until it is caught up or the worker stops; then a snapshot, and the group given back.
+fn run_group(api: &Arc<Api>, lease: GroupLeaseJson, lane: &str, config: &Config, stopping: &AtomicBool) {
+    let group_id = lease.group.id;
+    let name = lease.group.name.clone();
+    // The log's label; the indexer gets the group's name itself (keys, xrefs).
+    let group_name = if lane == "tail" { format!("{name} (tail)") } else { name.clone() };
+    let rules = lease.rules.rules();
+    let mut indexer = match lease.snapshot {
+        Some(s) => match api.get_snapshot(group_id, lane) {
+            Ok(Some((next, bytes))) => match RangeIndexer::restore(&bytes, rules.clone(), 5) {
+                Ok(ix) => {
+                    log("info", &format!("{group_name}: window restored from the snapshot at {next} ({:.1} MB).", bytes.len() as f64 / 1e6));
+                    ix
+                }
+                Err(e) => {
+                    log("warn", &format!("{group_name}: the snapshot at {} can't be restored ({e}); the window starts empty.", s.next));
+                    RangeIndexer::for_group(&name, rules.clone(), 5)
+                }
+            },
+            Ok(None) => {
+                log("warn", &format!("{group_name}: the lease names a snapshot at {} but there is none; the window starts empty.", s.next));
+                RangeIndexer::for_group(&name, rules.clone(), 5)
+            }
+            Err(e) => {
+                log("warn", &format!("{group_name}: snapshot: {e}; the group goes back."));
+                let _ = api.group_call(group_id, lane, "release");
+                return;
+            }
+        },
+        None => RangeIndexer::for_group(&name, rules.clone(), 5),
+    };
+    let totals = Arc::new(Mutex::new(Totals::default()));
+    let started = Instant::now();
+    let mut snapshotted = Instant::now();
+    let mut ranges = 0;
+    loop {
+        if stopping.load(Ordering::Relaxed) {
+            break;
+        }
+        let range = match api.next_range(group_id, lane) {
+            Ok(Some(range)) => range,
+            Ok(None) => {
+                log("info", &format!("{group_name}: caught up after {ranges} ranges; the group goes back."));
+                break;
+            }
+            Err(e) => {
+                log("warn", &format!("{group_name}: next range: {e}; the group goes back."));
+                break;
+            }
+        };
+        if range.next != indexer.next() && ranges > 0 {
+            log("warn", &format!("{group_name}: the window stands at {} but the next range resumes at {}: the window carries on from there.", indexer.next(), range.next));
+        }
+        match run_group_range(api, group_id, lane, &group_name, range, &mut indexer, config, stopping, &totals) {
+            RangeOutcome::Done => ranges += 1,
+            // The same range again from the reported chunk: the stream waits for its new copy.
+            RangeOutcome::Again => {}
+            RangeOutcome::Stop => break,
+        }
+        if snapshotted.elapsed() >= config.snapshot_every {
+            snapshot(api, group_id, lane, &group_name, &indexer);
+            snapshotted = Instant::now();
+        }
+    }
+    snapshot(api, group_id, lane, &group_name, &indexer);
+    let _ = api.group_call(group_id, lane, "release");
+    let t = totals.lock().unwrap();
+    log(
+        "info",
+        &format!(
+            "{group_name}: {ranges} ranges, {} headers in {:.0} s, {} release chunks, {} probe candidates; posted {:.1} MB in {:.1} s.",
+            t.headers,
+            started.elapsed().as_secs_f64(),
+            t.releases,
+            t.candidates,
+            t.bytes as f64 / 1e6,
+            t.post_ms as f64 / 1000.0
+        ),
+    );
+}
+
 fn main() {
     let args: Vec<String> = std::env::args().collect();
     let flag = |name: &str| args.iter().position(|a| a == name).and_then(|i| args.get(i + 1)).cloned();
@@ -520,9 +814,16 @@ fn main() {
         .filter(|&n| n > 0)
         .unwrap_or_else(|| std::thread::available_parallelism().map(|n| n.get()).unwrap_or(1));
     let ranges: Option<usize> = flag("--ranges").map(|v| v.parse().unwrap());
+    // Of the threads, this many on the groups' tail lane: one by default when there are two or more.
+    let tail_threads: usize = flag("--tail-threads")
+        .or_else(|| std::env::var("EDGE_TAIL_THREADS").ok())
+        .map(|v| v.parse().unwrap())
+        .unwrap_or(if threads >= 2 { 1 } else { 0 });
+    assert!(tail_threads <= threads, "--tail-threads {tail_threads} is more than the {threads} threads");
     let max_mbit: Option<f64> = flag("--max-mbit").or_else(|| std::env::var("EDGE_MAX_MBIT").ok()).map(|v| v.parse().unwrap()).filter(|&m: &f64| m > 0.0);
     let config = Arc::new(Config {
         batch_articles: flag("--batch-articles").map(|v| v.parse().unwrap()).unwrap_or(500_000),
+        snapshot_every: Duration::from_secs(60 * flag("--snapshot-minutes").or_else(|| std::env::var("EDGE_SNAPSHOT_MINUTES").ok()).map(|v| v.parse::<u64>().unwrap()).unwrap_or(30)),
         // The node's cap, shared by its threads.
         bytes_per_second: max_mbit.map(|m| m * 1e6 / 8.0 / threads as f64),
         smaller: Mutex::new(HashMap::new()),
@@ -531,22 +832,41 @@ fn main() {
     signal_hook::flag::register(signal_hook::consts::SIGTERM, stopping.clone()).unwrap();
     signal_hook::flag::register(signal_hook::consts::SIGINT, stopping.clone()).unwrap();
     let api = Arc::new(Api::new(&url, &token));
-    log("info", &format!("Starting {threads} worker threads{}.", max_mbit.map(|m| format!(" (download cap {m} Mbit/s)")).unwrap_or_default()));
+    log("info", &format!("Starting {threads} worker threads{}{}.", if range_leases_wanted(&ranges) { String::new() } else { format!(", {tail_threads} on the tail") }, max_mbit.map(|m| format!(" (download cap {m} Mbit/s)")).unwrap_or_default()));
+    let groups: Option<usize> = flag("--groups").map(|v| v.parse().unwrap());
+    let range_leases = ranges.is_some();
     let workers: Vec<_> = (0..threads)
-        .map(|_| {
+        .map(|i| {
             let (api, config, stopping) = (api.clone(), config.clone(), stopping.clone());
+            let lane = if i < tail_threads { "tail" } else { "history" };
             std::thread::spawn(move || {
                 let mut done = 0;
-                while !stopping.load(Ordering::Relaxed) && ranges.is_none_or(|n| done < n) {
-                    match api.lease() {
+                if range_leases {
+                    while !stopping.load(Ordering::Relaxed) && ranges.is_none_or(|n| done < n) {
+                        match api.lease() {
+                            Ok(Some(lease)) => {
+                                run_range(&api, lease, &config, &stopping);
+                                done += 1;
+                            }
+                            Ok(None) => break,
+                            Err(e) => {
+                                log("warn", &format!("Lease failed: {e}"));
+                                idle(&stopping);
+                            }
+                        }
+                    }
+                    return;
+                }
+                while !stopping.load(Ordering::Relaxed) && groups.is_none_or(|n| done < n) {
+                    match api.lease_group(lane) {
                         Ok(Some(lease)) => {
-                            run_range(&api, lease, &config, &stopping);
+                            run_group(&api, lease, lane, &config, &stopping);
                             done += 1;
                         }
-                        Ok(None) if ranges.is_some() => break,
+                        Ok(None) if groups.is_some() => break,
                         Ok(None) => idle(&stopping),
                         Err(e) => {
-                            log("warn", &format!("Lease failed: {e}"));
+                            log("warn", &format!("Group lease failed: {e}"));
                             idle(&stopping);
                         }
                     }
@@ -559,6 +879,10 @@ fn main() {
     }
 }
 
+fn range_leases_wanted(ranges: &Option<usize>) -> bool {
+    ranges.is_some()
+}
+
 fn idle(stopping: &AtomicBool) {
     let until = Instant::now() + IDLE;
     while Instant::now() < until && !stopping.load(Ordering::Relaxed) {

edge/src/lease.rs

@@ -0,0 +1,117 @@
+//! A lease as the coordinator sends it (lib/edge/lease.ts): the range and the rules, JSON with the TypeScript
+//! names. The edge-node worker takes one per range; edge-bench's replay takes the rules alone (`--rules`).
+
+use crate::names::Rules;
+use serde::Deserialize;
+use std::collections::HashMap;
+
+#[derive(Deserialize)]
+pub struct LeaseJson {
+    pub range: RangeJson,
+    pub rules: RulesJson,
+}
+
+/// A group lease (lib/edge/lease.ts GroupLease): the group a worker thread holds, the rules, the window's snapshot if any.
+#[derive(Deserialize)]
+pub struct GroupLeaseJson {
+    pub group: GroupJson,
+    pub rules: RulesJson,
+    pub snapshot: Option<SnapshotJson>,
+}
+
+#[derive(Deserialize)]
+pub struct GroupJson {
+    pub id: i64,
+    pub name: String,
+    pub priority: i64,
+}
+
+#[derive(Deserialize)]
+pub struct SnapshotJson {
+    pub next: i64,
+}
+
+/// The group's next range (lib/edge/lease.ts NextRange).
+#[derive(Deserialize)]
+pub struct NextRangeJson {
+    pub range: RangeSpanJson,
+}
+
+#[derive(Deserialize, Clone, Copy)]
+pub struct RangeSpanJson {
+    pub id: i64,
+    pub start: i64,
+    pub end: i64,
+    pub next: i64,
+}
+
+#[derive(Deserialize)]
+pub struct RangeJson {
+    pub id: i64,
+    pub group: String,
+    pub start: i64,
+    pub end: i64,
+    pub next: i64,
+}
+
+#[derive(Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct RulesJson {
+    pub version: i64,
+    pub priorities: HashMap<String, u32>,
+    #[serde(default)]
+    pub dropped_signatures: Vec<String>,
+    #[serde(default)]
+    pub dropped_classes: Vec<String>,
+    #[serde(default)]
+    pub dropped_tools: Option<Vec<String>>,
+    #[serde(default)]
+    pub per_article_max_parts: Option<u32>,
+    #[serde(default)]
+    pub sample_percent: Option<f64>,
+    #[serde(default)]
+    pub drop_executables: Option<bool>,
+    #[serde(default)]
+    pub dropped_posters: Option<Vec<String>>,
+    #[serde(default)]
+    pub drop_stub_archives: Option<bool>,
+    #[serde(default)]
+    pub drop_nonsense_pieces: Option<bool>,
+    #[serde(default)]
+    pub drop_tiny_archives: Option<bool>,
+}
+
+impl RulesJson {
+    /// The rules with today's defaults under them (a field the stored set doesn't have keeps the default).
+    pub fn rules(self) -> Rules {
+        let mut rules = Rules::defaults(self.priorities.into_iter().collect());
+        rules.version = self.version;
+        rules.dropped_signatures = self.dropped_signatures.into_iter().collect();
+        rules.dropped_classes = self.dropped_classes.into_iter().collect();
+        if let Some(tools) = self.dropped_tools {
+            rules.dropped_tools = tools.into_iter().collect();
+        }
+        if let Some(n) = self.per_article_max_parts {
+            rules.per_article_max_parts = n;
+        }
+        if let Some(p) = self.sample_percent {
+            rules.sample_percent = p;
+        }
+        if let Some(e) = self.drop_executables {
+            rules.drop_executables = e;
+        }
+        if let Some(posters) = self.dropped_posters {
+            rules.dropped_posters = posters;
+        }
+        if let Some(d) = self.drop_stub_archives {
+            rules.drop_stub_archives = d;
+        }
+        if let Some(d) = self.drop_nonsense_pieces {
+            rules.drop_nonsense_pieces = d;
+        }
+        if let Some(d) = self.drop_tiny_archives {
+            rules.drop_tiny_archives = d;
+        }
+        rules
+    }
+}

edge/src/lib.rs

@@ -5,5 +5,6 @@
 pub mod codec;
 pub mod indexer;
 pub mod js;
+pub mod lease;
 pub mod names;
 pub mod wire;

edge/src/main.rs

@@ -1,10 +1,11 @@
-//! The edge worker's indexing in Rust, on a range captured by `edge:bench
-//! --capture` (REBUILD.md, part 2: TypeScript or Rust). Same work as
-//! edge:bench's replay: inflate each chunk, scan it, aggregate files, cut off
-//! idle ones, and every N chunks judge and encode the releases. `--dump`
-//! writes every output as JSON lines for edge/check.ts.
+//! The Rust indexer over stored chunks (REBUILD.md, part 2 and "Voyager 3"): a range captured by `edge:bench
+//! --capture` or stored in the corpus, or a group's corpus ranges in order through one window (`--group`, the
+//! replay harness behind `replay:corpus`). Same work as the edge-node worker: inflate each chunk, scan it,
+//! aggregate files, finish releases through the window, and every N chunks judge and encode them. `--dump`
+//! writes every output as JSON lines; `--wire` the batches as the worker posts them (`-`: stdout).
 //!
-//!   edge-bench <dir> [--batch-chunks 10] [--idle 5] [--dump out.jsonl] [--wire batches.jsonl]
+//!   edge-bench <range dir> | --group <group dir> [--from article] [--to article] [--limit ranges] [--rules rules.json]
+//!              [--batch-chunks 10] [--idle 5] [--no-finish] [--dump out.jsonl] [--wire batches.jsonl|-] [--releases out.jsonl]
 
 use base64::Engine;
 use edge::indexer::{self, Batch, RangeIndexer};
@@ -123,7 +124,7 @@ fn dump_batch(batch: &Batch, out: &mut impl Write) {
         line.push_str(",\"postedLast\":");
         opt_num(r.posted_last, line);
         line.push_str(&format!(
-            ",\"files\":{},\"segments\":{},\"bytes\":{},\"firstArticle\":{},\"lastArticle\":{},\"readable\":{},\"hasPar2\":{},\"payload\":\"{}\"",
+            ",\"files\":{},\"segments\":{},\"bytes\":{},\"firstArticle\":{},\"lastArticle\":{},\"readable\":{},\"hasPar2\":{},\"finished\":{},\"edge\":{},\"partial\":{},\"payload\":\"{}\"",
             r.files,
             r.segments,
             r.bytes,
@@ -131,6 +132,9 @@ fn dump_batch(batch: &Batch, out: &mut impl Write) {
             r.last_article,
             r.readable,
             r.has_par2,
+            r.finished,
+            r.edge,
+            r.partial,
             b64.encode(&r.payload)
         ));
     };
@@ -162,122 +166,220 @@ fn dump_batch(batch: &Batch, out: &mut impl Write) {
         line.push_str(&format!(",\"articles\":{articles},\"files\":{files}}}"));
         emit(&mut line);
     }
-    for f in &batch.fragments {
-        line.push_str("{\"t\":\"f\",\"fileKey\":");
-        json_str(&f.file_key, &mut line);
-        line.push_str(",\"name\":");
-        json_str(&f.name, &mut line);
-        line.push_str(&format!(",\"totalParts\":{},\"postedAt\":", f.total_parts));
-        opt_num(f.posted_at, &mut line);
-        line.push_str(",\"segments\":[");
-        for (i, (n, id, bytes, article)) in f.segments.iter().enumerate() {
-            if i > 0 {
-                line.push(',');
-            }
-            line.push_str(&format!("[{n},"));
-            json_str(id, &mut line);
-            line.push_str(&format!(",{bytes},{article}]"));
+}
+
+/// A stored range: its directory and article span (range.json).
+struct Range {
+    dir: String,
+    start: i64,
+    end: i64,
+}
+
+/// The range in a directory with a range.json, and its group.
+fn range_in(dir: &str) -> Option<(String, Range)> {
+    let json = std::fs::read_to_string(format!("{dir}/range.json")).ok()?;
+    Some((json_field(&json, "group"), Range { dir: dir.to_string(), start: json_field(&json, "start").parse().ok()?, end: json_field(&json, "end").parse().ok()? }))
+}
+
+/// A group directory's complete ranges in article order (the corpus layout: `<group>/<start>-<end>/`).
+fn ranges_in(group_dir: &str) -> (String, Vec<Range>) {
+    let mut group = String::new();
+    let mut ranges: Vec<Range> = Vec::new();
+    for entry in std::fs::read_dir(group_dir).unwrap_or_else(|e| panic!("{group_dir}: {e}")) {
+        let path = entry.unwrap().path();
+        if !path.is_dir() {
+            continue;
         }
-        line.push_str("]}");
-        emit(&mut line);
+        if let Some((g, range)) = range_in(path.to_str().unwrap()) {
+            group = g;
+            ranges.push(range);
+        }
+    }
+    ranges.sort_by_key(|r| r.start);
+    (group, ranges)
+}
+
+fn date(ms: Option<i64>) -> String {
+    match ms.and_then(chrono::DateTime::from_timestamp_millis) {
+        Some(t) => t.format("%Y-%m-%d %H:%M").to_string(),
+        None => "-".to_string(),
     }
 }
 
+/// Files waiting in finished releases after which a batch goes out whatever its chunks (edge-node's limit).
+const MAX_PENDING_FILES: usize = 100_000;
+
 fn main() {
     let args: Vec<String> = std::env::args().collect();
-    let dir = args.get(1).expect("usage: edge-bench <dir> [--batch-chunks N] [--idle N] [--dump file]").clone();
+    let usage = "usage: edge-bench <range dir> | --group <group dir> [--from article] [--to article] [--limit ranges] [--rules rules.json] [--batch-chunks N] [--idle N] [--no-finish] [--dump file] [--wire file|-] [--releases file]";
     let flag = |name: &str| args.iter().position(|a| a == name).and_then(|i| args.get(i + 1)).cloned();
+    let has = |name: &str| args.iter().any(|a| a == name);
     let batch_chunks: usize = flag("--batch-chunks").map(|v| v.parse().unwrap()).unwrap_or(10);
     let idle: u64 = flag("--idle").map(|v| v.parse().unwrap()).unwrap_or(5);
     let mut dump = flag("--dump").map(|p| std::io::BufWriter::new(std::fs::File::create(p).unwrap()));
-    // Each batch as the worker posts it (wire.rs), one JSON per line: check.ts compares it with toWire.
-    let mut wire_out = flag("--wire").map(|p| std::io::BufWriter::new(std::fs::File::create(p).unwrap()));
+    // Each batch as the worker posts it (wire.rs), one JSON per line: a file, or stdout (`-`) for replay:corpus.
+    let wire_path = flag("--wire");
+    let wire_stdout = wire_path.as_deref() == Some("-");
+    let mut wire_out: Option<Box<dyn Write>> = match wire_path.as_deref() {
+        Some("-") => Some(Box::new(std::io::BufWriter::with_capacity(1 << 20, std::io::stdout().lock()))),
+        Some(p) => Some(Box::new(std::io::BufWriter::new(std::fs::File::create(p).unwrap()))),
+        None => None,
+    };
     // One line per release chunk without its payload, with its first subjects (deploy/corpus_sweep.py).
     let mut releases_out = flag("--releases").map(|p| std::io::BufWriter::new(std::fs::File::create(p).unwrap()));
 
-    let range = std::fs::read_to_string(format!("{dir}/range.json")).unwrap();
-    let group = json_field(&range, "group");
-    let start: i64 = json_field(&range, "start").parse().unwrap();
-    let end: i64 = json_field(&range, "end").parse().unwrap();
-    let mut files: Vec<String> = std::fs::read_dir(&dir).unwrap().map(|e| e.unwrap().file_name().into_string().unwrap()).filter(|f| f.ends_with(".gz")).collect();
-    files.sort();
+    // One range, or a group's ranges in order through one window (the window carries over from range to range;
+    // the end is only finished after the last one, unless --no-finish: a replay to be continued).
+    let (group, ranges) = match flag("--group") {
+        Some(dir) => {
+            let (group, mut ranges) = ranges_in(&dir);
+            // --from: ranges ending before it are skipped (a replay continued); --to: ranges starting after it (one
+            // window of the group's history, the corpus holding others); --limit: this many at most.
+            if let Some(from) = flag("--from").map(|v| v.parse::<i64>().unwrap()) {
+                ranges.retain(|r| r.end >= from);
+            }
+            if let Some(to) = flag("--to").map(|v| v.parse::<i64>().unwrap()) {
+                ranges.retain(|r| r.start <= to);
+            }
+            if let Some(limit) = flag("--limit").map(|v| v.parse::<usize>().unwrap()) {
+                ranges.truncate(limit);
+            }
+            (group, ranges)
+        }
+        None => {
+            let positional = (1..args.len()).find(|&i| !args[i].starts_with("--") && !args[i - 1].starts_with("--")).map(|i| args[i].clone());
+            let dir = positional.expect(usage);
+            let (group, range) = range_in(&dir).unwrap_or_else(|| panic!("{dir}: no range.json"));
+            (group, vec![range])
+        }
+    };
+    if ranges.is_empty() {
+        eprintln!("no complete range found");
+        std::process::exit(2);
+    }
+    let finish = !has("--no-finish");
+    let rules = match flag("--rules") {
+        Some(path) => {
+            let json = std::fs::read_to_string(&path).unwrap_or_else(|e| panic!("{path}: {e}"));
+            serde_json::from_str::<edge::lease::RulesJson>(&json).unwrap_or_else(|e| panic!("{path}: {e}")).rules()
+        }
+        None => {
+            let mut priorities = FxHashMap::default();
+            priorities.insert(group.clone(), 1);
+            names::Rules::defaults(priorities)
+        }
+    };
+    // The report goes to stderr when the wire takes stdout.
+    let say = |line: String| {
+        if wire_stdout {
+            eprintln!("{line}");
+        } else {
+            println!("{line}");
+        }
+    };
 
-    let mut priorities = FxHashMap::default();
-    priorities.insert(group.clone(), 1);
-    let mut indexer = RangeIndexer::new(&group, start, end, start, names::Rules::defaults(priorities), idle);
+    let (start, end) = (ranges[0].start, ranges[ranges.len() - 1].end);
+    let mut indexer = RangeIndexer::new(&group, start, end, start, rules, idle);
 
     let (mut inflate_s, mut index_s) = (0.0, 0.0);
-    let (mut compressed, mut text_bytes) = (0usize, 0usize);
-    let mut out = (0u64, 0usize, 0usize, 0usize, 0usize, 0usize, 0usize, 0usize); // headers, releases, files, segments, payload, candidates, fragments, samples
+    let (mut compressed, mut text_bytes, mut chunks) = (0usize, 0usize, 0usize);
+    let mut out = (0u64, 0usize, 0usize, 0usize, 0usize, 0usize, 0usize); // headers, releases, files, segments, payload, candidates, samples
+    let mut finished = 0usize;
     let mut keys = std::collections::HashSet::new();
     let mut decompressor = libdeflater::Decompressor::new();
     let wall = Instant::now();
-    for (i, file) in files.iter().enumerate() {
-        let raw = std::fs::read(format!("{dir}/{file}")).unwrap();
-        compressed += raw.len();
-        let t = Instant::now();
-        let data = inflate(&raw, &mut decompressor);
-        inflate_s += t.elapsed().as_secs_f64();
-        text_bytes += data.len();
-        let t = Instant::now();
-        let text = String::from_utf8_lossy(&data);
-        let next: i64 = file.split('-').nth(2).unwrap().trim_end_matches(".gz").parse::<i64>().unwrap() + 1;
-        indexer.add_text(&text, next);
-        let last = i + 1 == files.len();
-        if last {
-            indexer.finish();
+    let mut since_batch = 0usize;
+    for (ri, range) in ranges.iter().enumerate() {
+        let range_wall = Instant::now();
+        let headers_before = out.0;
+        // After a gap between stored ranges the batch starts at this range (every range ends with a batch taken).
+        if ri > 0 {
+            indexer.skip_to(range.start);
         }
-        if !last && (i + 1) % batch_chunks != 0 {
+        let mut files: Vec<String> = std::fs::read_dir(&range.dir).unwrap().map(|e| e.unwrap().file_name().into_string().unwrap()).filter(|f| f.ends_with(".gz")).collect();
+        files.sort();
+        for (i, file) in files.iter().enumerate() {
+            let raw = std::fs::read(format!("{}/{file}", range.dir)).unwrap();
+            compressed += raw.len();
+            let t = Instant::now();
+            let data = inflate(&raw, &mut decompressor);
+            inflate_s += t.elapsed().as_secs_f64();
+            text_bytes += data.len();
+            chunks += 1;
+            let t = Instant::now();
+            let text = String::from_utf8_lossy(&data);
+            let next: i64 = file.split('-').nth(2).unwrap().trim_end_matches(".gz").parse::<i64>().unwrap() + 1;
+            indexer.add_text(&text, next);
+            since_batch += 1;
+            let range_end = i + 1 == files.len();
+            let last = range_end && ri + 1 == ranges.len();
+            if last && finish {
+                indexer.finish();
+            }
+            if !range_end && since_batch < batch_chunks && indexer.pending() < MAX_PENDING_FILES {
+                index_s += t.elapsed().as_secs_f64();
+                continue;
+            }
+            let batch = indexer.take();
+            since_batch = 0;
             index_s += t.elapsed().as_secs_f64();
-            continue;
+            out.0 += batch.headers;
+            out.1 += batch.releases.len();
+            for r in &batch.releases {
+                keys.insert(r.key.clone());
+                out.2 += r.files;
+                out.3 += r.segments;
+                out.4 += r.payload.len();
+                finished += r.finished as usize;
+            }
+            out.5 += batch.candidates.len();
+            out.6 += batch.samples.len();
+            if let Some(d) = dump.as_mut() {
+                dump_batch(&batch, d);
+            }
+            if let Some(w) = wire_out.as_mut() {
+                w.write_all(&wire::to_json(&batch)).unwrap();
+                w.write_all(b"\n").unwrap();
+                w.flush().unwrap();
+            }
+            if let Some(w) = releases_out.as_mut() {
+                dump_releases(&batch, &group, w);
+            }
         }
-        let batch = indexer.take();
-        index_s += t.elapsed().as_secs_f64();
-        out.0 += batch.headers;
-        out.1 += batch.releases.len();
-        for r in &batch.releases {
-            keys.insert(r.key.clone());
-            out.2 += r.files;
-            out.3 += r.segments;
-            out.4 += r.payload.len();
-        }
-        out.5 += batch.candidates.len();
-        out.6 += batch.fragments.len();
-        out.7 += batch.samples.len();
-        if let Some(d) = dump.as_mut() {
-            dump_batch(&batch, d);
-        }
-        if let Some(w) = wire_out.as_mut() {
-            w.write_all(&wire::to_json(&batch)).unwrap();
-            w.write_all(b"\n").unwrap();
-        }
-        if let Some(w) = releases_out.as_mut() {
-            dump_releases(&batch, &group, w);
+        if ranges.len() > 1 {
+            eprintln!(
+                "{} {}–{}: {} headers in {:.1} s; frontier {}; {} releases out ({} finished), {} releases and {} files open; RSS {} MB",
+                group,
+                range.start,
+                range.end,
+                out.0 - headers_before,
+                range_wall.elapsed().as_secs_f64(),
+                date(indexer.frontier()),
+                out.1,
+                finished,
+                indexer.open_releases(),
+                indexer.open_files(),
+                peak_rss_mb()
+            );
         }
     }
     let total = inflate_s + index_s;
     #[allow(static_mut_refs)]
     let p = unsafe { indexer::PROF };
     eprintln!("  of which aggregate {:.2} s, evict+cutoff {:.2} s, judge+encode {:.2} s", p[3] as f64 / 1e9, p[4] as f64 / 1e9, p[5] as f64 / 1e9);
-    println!("{group} {start}–{end}: {} chunks, {} MB compressed, {} MB text", files.len(), compressed / 1_000_000, text_bytes / 1_000_000);
-    println!(
-        "{} headers in {:.2} s ({:.0}/s on one core; wall {:.2} s)",
-        out.0,
-        total,
-        out.0 as f64 / total,
-        wall.elapsed().as_secs_f64()
-    );
-    println!("  inflate {inflate_s:.2} s, index {index_s:.2} s; peak RSS {} MB", peak_rss_mb());
-    println!(
-        "  kept: {} release chunks ({} releases), {} files, {} segments, payload {:.1} MB ({:.1} B/segment); {} candidates, {} fragments, {} samples",
+    say(format!("{group} {start}–{end}: {} ranges, {chunks} chunks, {} MB compressed, {} MB text", ranges.len(), compressed / 1_000_000, text_bytes / 1_000_000));
+    say(format!("{} headers in {:.2} s ({:.0}/s on one core; wall {:.2} s)", out.0, total, out.0 as f64 / total, wall.elapsed().as_secs_f64()));
+    say(format!("  inflate {inflate_s:.2} s, index {index_s:.2} s; peak RSS {} MB", peak_rss_mb()));
+    say(format!(
+        "  kept: {} release chunks ({} releases, {} finished), {} files, {} segments, payload {:.1} MB ({:.1} B/segment); {} candidates, {} samples",
         out.1,
         keys.len(),
+        finished,
         out.2,
         out.3,
         out.4 as f64 / 1e6,
         out.4 as f64 / out.3.max(1) as f64,
         out.5,
-        out.6,
-        out.7
-    );
+        out.6
+    ));
 }

edge/src/names.rs

@@ -449,6 +449,43 @@ fn finish_remainder(s: &str) -> String {
     js::trim(&out).to_string()
 }
 
+/// displayName: a subject name as shown, brackets the counters left unmatched removed ("… Facialized 002/477] - …"
+/// gave "… Facialized ]"). Only the name: the key keeps the remainder.
+pub fn display_name(remainder: &str) -> String {
+    let b = remainder.as_bytes();
+    if !b.iter().any(|c| matches!(c, b'[' | b']' | b'(' | b')')) {
+        return remainder.to_string();
+    }
+    let mut drop = vec![false; b.len()];
+    for (open, close) in [(b'[', b']'), (b'(', b')')] {
+        let mut opens: Vec<usize> = Vec::new();
+        for (i, &c) in b.iter().enumerate() {
+            if c == open {
+                opens.push(i);
+            } else if c == close && opens.pop().is_none() {
+                drop[i] = true;
+            }
+        }
+        for i in opens {
+            drop[i] = true;
+        }
+    }
+    if !drop.contains(&true) {
+        return remainder.to_string();
+    }
+    let kept: Vec<u8> = b.iter().zip(&drop).filter(|(_, &d)| !d).map(|(&c, _)| c).collect();
+    let kept = String::from_utf8(kept).expect("only ASCII brackets removed");
+    let mut out = String::with_capacity(kept.len());
+    for c in kept.chars() {
+        if c == ' ' && out.ends_with(' ') {
+            continue;
+        }
+        out.push(c);
+    }
+    let name = js::trim(&out);
+    if name.is_empty() { remainder.to_string() } else { name.to_string() }
+}
+
 /// counterTotal: the subject's file counter total, first form found first.
 pub fn counter_total(subject: &str) -> Option<String> {
     let s = replace(&QUOTED, subject, " ", true);
@@ -841,11 +878,11 @@ pub fn release_identity(poster: &str, subject: &str, name: &str) -> Release {
         if !image && test(&RANDOM_VOLUME, name) && stem.bytes().all(|c| c.is_ascii_alphanumeric() || c == b'_' || c == b'-') && looks_random(&stripped) {
             parts.push(&stem);
         }
-        return Release { key: hash(&parts), name: remainder, source: "subject", file_total };
+        return Release { key: hash(&parts), name: display_name(&remainder), source: "subject", file_total };
     }
     if let Some(total) = post_total.filter(|&n| n != 0 && !tag) {
         if is_post_token(&remainder) {
-            return Release { key: hash(&["post", poster, &remainder, &total.to_string()]), name: remainder, source: "post", file_total };
+            return Release { key: hash(&["post", poster, &remainder, &total.to_string()]), name: display_name(&remainder), source: "post", file_total };
         }
     }
     // A scene music post's index files ("00-ammar-turk_uit_de_kroeg-single-web-nl-2022-cover.jpg") under a bare counter go
@@ -1028,6 +1065,228 @@ pub fn is_random_file(name: &str) -> bool {
     looks_random(&file_stem(name))
 }
 
+/// Readable or not (lib/release/readable.ts): whether a name says what a release is, read token by token, so
+/// separators, spaces and random "extensions" don't hide a scramble that looks_random's single-token shapes miss
+/// (LOG.md (129)). The readable flag; the drop rules keep looks_random, which is narrower.
+/// Tokens that don't make a name readable on their own: packaging, volume numbers and the posting tools' fixed words.
+static NOISE_TOKEN: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)^(addon|samples?|proof|subs?|nfo|part[0-9]+|vol[0-9]+|description|repost|rep|extras?|bonus|cd[0-9]*|dvd[0-9]*|dis[ck][0-9]*|par2|rar|zip|7z|mkv|mp4|avi|vol)$"));
+/// Mixed tokens that are codes, not scrambles: S02E10, 1080p, 10bit, x264, Win10x64, BDRip720p, 3DMark03, v2, 5th.
+static CODE_TOKEN: LazyLock<Regex> = LazyLock::new(|| re(r"^([sS][0-9]{1,2}[eE][0-9]{1,3}|[xh]26[45]|[A-Za-z]+[0-9]+([a-z]{1,3}|x(64|86))?|[0-9]+[A-Za-z]{1,3}|[0-9]+[A-Za-z]+[0-9]+)$"));
+/// The NZBSPLIT tools' hash between their markers is no part of the name.
+static NZBSPLIT_HASH: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)__NZBS[P]LIT__[0-9a-f]{32}__NZBS[P]LIT__")); // "SP" is re()'s space class
+/// An inner run of 2+ capitals with lowercase on both sides ("tBWTuz", "YDoaROv"; not "3DMark03").
+static INNER_UPPER: LazyLock<Regex> = LazyLock::new(|| re(r"[a-z][A-Z]{2,}[a-z]"));
+
+/// Scene tags write "i" for "I" between capitals (XCiTE, PRESCRiPTiON): read it as a capital there.
+fn scene_case(t: &[u8]) -> Vec<u8> {
+    let mut out = t.to_vec();
+    for i in 1..t.len().saturating_sub(1) {
+        if t[i] == b'i' && t[i - 1].is_ascii_uppercase() && t[i + 1].is_ascii_uppercase() {
+            out[i] = b'I';
+        }
+    }
+    out
+}
+
+/// The signals of a scrambled token (readable.ts signals()).
+struct TokenSignals {
+    /// Letter/digit and lower-to-upper switches.
+    switches: usize,
+    /// Case switches in both directions.
+    case_switches: usize,
+    /// Lone lowercase letters (not "i") between capitals ("NtpUgCmS").
+    lone: usize,
+    inner_upper: bool,
+    max_consonants: usize,
+    /// Vowels, a lowercase y too ("McCartney").
+    vowels: usize,
+    /// Runs of one character class.
+    runs: usize,
+    letters: usize,
+    /// The longest lowercase run.
+    max_lower: usize,
+}
+
+fn token_signals(t: &[u8]) -> TokenSignals {
+    let is_vowel = |c: u8| matches!(c.to_ascii_lowercase(), b'a' | b'e' | b'i' | b'o' | b'u') || c == b'y';
+    let class = |c: u8| if c.is_ascii_digit() { 0 } else if c.is_ascii_lowercase() { 1 } else { 2 };
+    let mut s = TokenSignals { switches: 0, case_switches: 0, lone: 0, inner_upper: false, max_consonants: 0, vowels: 0, runs: 0, letters: 0, max_lower: 0 };
+    let mut consonants = 0;
+    let mut lower = 0;
+    for i in 0..t.len() {
+        let c = t[i];
+        if c.is_ascii_alphabetic() && !is_vowel(c) && c.to_ascii_lowercase() != b'y' {
+            consonants += 1;
+            s.max_consonants = s.max_consonants.max(consonants);
+        } else {
+            consonants = 0;
+        }
+        if is_vowel(c) {
+            s.vowels += 1;
+        }
+        if c.is_ascii_alphabetic() {
+            s.letters += 1;
+        }
+        if c.is_ascii_lowercase() {
+            lower += 1;
+            s.max_lower = s.max_lower.max(lower);
+        } else {
+            lower = 0;
+        }
+        if i == 0 || class(t[i - 1]) != class(c) {
+            s.runs += 1;
+        }
+        if i == 0 {
+            continue;
+        }
+        let a = t[i - 1];
+        if a.is_ascii_digit() != c.is_ascii_digit() {
+            s.switches += 1;
+        } else if a.is_ascii_lowercase() && c.is_ascii_uppercase() {
+            s.switches += 1;
+        }
+        if (a.is_ascii_lowercase() && c.is_ascii_uppercase()) || (a.is_ascii_uppercase() && c.is_ascii_lowercase()) {
+            s.case_switches += 1;
+        }
+        if i < t.len() - 1 && c.is_ascii_lowercase() && c != b'i' && a.is_ascii_uppercase() && t[i + 1].is_ascii_uppercase() {
+            s.lone += 1;
+        }
+    }
+    s.inner_upper = INNER_UPPER.is_match(t);
+    s
+}
+
+#[derive(PartialEq, Eq, Debug, Clone, Copy)]
+pub enum TokenKind {
+    Word,
+    Random,
+    Neutral,
+}
+
+/// What one alphanumeric token of a name is: a word (letters only, in one case or Capitalized, 3+ long, not a
+/// packaging word), a scramble, or neither (digits, codes, CamelCase, short tokens). Kept in step with readable.ts.
+pub fn token_kind(t: &[u8]) -> TokenKind {
+    let n = t.len();
+    if t.iter().all(u8::is_ascii_digit) || NOISE_TOKEN.is_match(t) || CODE_TOKEN.is_match(t) {
+        return TokenKind::Neutral;
+    }
+    if t.iter().all(u8::is_ascii_alphabetic) {
+        let s = token_signals(&scene_case(t));
+        let capitalized = t[0].is_ascii_uppercase() && t[1..].iter().all(u8::is_ascii_lowercase);
+        if s.case_switches == 0 || (s.case_switches == 1 && capitalized) {
+            // A base32 consonant soup ("mrrtmzrzmi4dgmrzmyytqmlbztczbx" without its digit) is no word.
+            if n >= 10 && s.vowels * 6 < n {
+                return TokenKind::Random;
+            }
+            return if n >= 3 { TokenKind::Word } else { TokenKind::Neutral };
+        }
+        // Mixed case beyond Capitalization: CamelCase (neutral) or a scramble, which has few vowels and short runs.
+        // Two CamelCase words with few vowels (FastFilms, BestCrypt, SchiffsModell) have four runs and no lone
+        // letter; a scramble has more runs, a lone letter or an inner capital run.
+        if n >= 6 && s.case_switches >= 2 && s.vowels * 4 < n && (s.runs >= 5 || s.lone >= 1 || s.inner_upper) {
+            return TokenKind::Random;
+        }
+        return TokenKind::Neutral;
+    }
+    if n >= 8 && t.iter().all(u8::is_ascii_hexdigit) && t.iter().filter(|c| c.is_ascii_digit()).count() >= 2 {
+        return TokenKind::Random;
+    }
+    if t.iter().all(|c| c.is_ascii_uppercase() || c.is_ascii_digit()) {
+        // 51294P1CUKR2R35X; COD4MW2REMASTERED (4 switches) and BEATLES1964REMASTER (2) stay.
+        return if n >= 12 && token_signals(t).switches >= 6 { TokenKind::Random } else { TokenKind::Neutral };
+    }
+    // Words with numbers between them (Laura15Juli2026, Back4WinXP): every letter run a word or an acronym.
+    if t.split(|c: &u8| !c.is_ascii_alphabetic()).filter(|w| !w.is_empty()).all(|w| {
+        let vowels = w.iter().filter(|c| matches!(c.to_ascii_lowercase(), b'a' | b'e' | b'i' | b'o' | b'u' | b'y')).count();
+        (w.len() >= 2 && w.iter().all(u8::is_ascii_uppercase))
+            || (w.len() >= 2 && w[1..].iter().all(u8::is_ascii_lowercase) && (w.len() >= 3 || w[0].is_ascii_lowercase()) && vowels * 5 >= w.len())
+    }) {
+        return TokenKind::Neutral;
+    }
+    let s = token_signals(&scene_case(t));
+    if n >= 5 && s.switches >= 3 {
+        return TokenKind::Random;
+    }
+    if n >= 6 && s.switches >= 2 && (s.lone >= 1 || s.inner_upper || s.max_consonants >= 4) {
+        return TokenKind::Random;
+    }
+    // Few vowels and no lowercase run of 3+ ("Jk5Pud", "RKi9BCg"; not "HDR10Plus").
+    if n >= 6 && s.switches >= 2 && s.vowels * 4 < s.letters && s.max_lower <= 2 {
+        return TokenKind::Random;
+    }
+    TokenKind::Neutral
+}
+
+/// A file or release name no one can read (readable.ts isUnreadableName): looks_random's shapes, or more scrambled
+/// tokens than words by letter count ("Zx5AjNaK-mh7CFg6Hf8uu", "Description - Lt5nHjlo3RiuBb3C", "PsNdnYiugfZ"; not
+/// "Paul McCartney - Band on the Run", "ShesSoWild - Ester Ana 1080p").
+pub fn is_unreadable_name(name: &str) -> bool {
+    let stem = file_stem(name);
+    let stem = NZBSPLIT_HASH.replace(stem.as_bytes(), &b" "[..]);
+    if looks_random(&String::from_utf8_lossy(&stem)) {
+        return true;
+    }
+    let mut words = 0;
+    let mut random = 0;
+    for t in stem.split(|c| !c.is_ascii_alphanumeric()).filter(|t| !t.is_empty()) {
+        match token_kind(t) {
+            TokenKind::Word => words += t.len(),
+            TokenKind::Random => random += t.len(),
+            TokenKind::Neutral => {}
+        }
+    }
+    random > 0 && random > words
+}
+
+/// The class of an unnamed release (lib/release/classes.ts pendingClass; LOG.md (130)): size bucket × name shape ×
+/// partial, 1..70. The lease's dropped classes ("tool\tgroup\tklass") are skipped at the edge.
+pub fn pending_class(name: &str, bytes: u64, partial: bool) -> u8 {
+    let size = [1_000_000u64, 10_000_000, 100_000_000, 1_000_000_000].iter().filter(|&&b| bytes >= b).count() as u8;
+    let s = match name.rfind('.') {
+        Some(i) if (1..=4).contains(&(name.len() - i - 1)) && name[i + 1..].bytes().all(|b| b.is_ascii_alphanumeric()) => &name[..i],
+        _ => name,
+    };
+    let b = s.as_bytes();
+    let all = |f: fn(&u8) -> bool| !b.is_empty() && b.iter().all(f);
+    let shape: u8 = if b.len() >= 16 && all(u8::is_ascii_hexdigit) {
+        1
+    } else if b.len() >= 6 && all(u8::is_ascii_digit) {
+        2
+    } else if b.len() >= 20 && all(|c| c.is_ascii_lowercase() || (b'2'..=b'7').contains(c)) && b.iter().any(|c| (b'2'..=b'7').contains(c)) {
+        3
+    } else if all(u8::is_ascii_alphabetic) {
+        4
+    } else if all(u8::is_ascii_alphanumeric) {
+        5
+    } else if all(|c| c.is_ascii_alphanumeric() || matches!(c, b'.' | b'_' | b'-')) {
+        6
+    } else {
+        7
+    };
+    size * 14 + (shape - 1) * 2 + u8::from(partial) + 1
+}
+
+/// A class as droppedClasses lists it.
+pub fn class_key(tool: &str, group: &str, klass: u8) -> String {
+    format!("{tool}\t{group}\t{klass}")
+}
+
+/// looksScrambled: at least half the letters and digits of a file's stem in tokens of 5 or more that switch case
+/// (lower to upper) or between letters and digits 3 times or more ("nYDeXeF49dieZ1J7flK-JFzZ3j.ZEDsr", "bslXlwgPiqZadhRa").
+/// The names a random-subject-per-article upload gives each article that looksRandom misses for their separators.
+pub fn looks_scrambled(name: &str) -> bool {
+    let stem = file_stem(name);
+    let (mut total, mut scrambled) = (0, 0);
+    for token in stem.as_bytes().split(|c| !c.is_ascii_alphanumeric()).filter(|t| !t.is_empty()) {
+        total += token.len();
+        let switches = token.windows(2).filter(|w| (w[0].is_ascii_lowercase() && w[1].is_ascii_uppercase()) || w[0].is_ascii_digit() != w[1].is_ascii_digit()).count();
+        if token.len() >= 5 && switches >= 3 {
+            scrambled += token.len();
+        }
+    }
+    total >= 8 && scrambled * 2 >= total
+}
+
 static SEVEN_SPLIT: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)\.7z\.([0-9]{1,4})$"));
 static SEVEN: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)\.7z$"));
 static PART1: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)\.part0*1\.rar$"));
@@ -1237,7 +1496,45 @@ pub fn explored(key: &str, percent: f64) -> bool {
 }
 
 static SIDECAR: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)\.(par2|nfo|sfv|txt|jpe?g|png|nzb)$"));
+
+/// A sidecar by its extension (the judge's SIDECAR): no release of its own, and no album's member.
+pub fn is_sidecar(name: &str) -> bool {
+    test(&SIDECAR, name)
+}
 static EXECUTABLE: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)\.(exe|scr|com|pif|bat|cmd|vbs|lnk)SP*$"));
+/// An unsplit archive (the judge's ARCHIVE and VOLUME): one zip, rar or 7z that is not a volume of a set.
+static ARCHIVE: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)\.(zip|rar|7z)$"));
+/// "cneg" is "part" in ROT13: the a.b.erotica posts named that way are volumes of passworded videos (LOG.md (128)).
+static VOLUME: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)\.((part|cneg)[0-9]+\.rar|r[0-9][0-9]|z[0-9][0-9]|7z\.[0-9][0-9][0-9]|[0-9][0-9][0-9])$"));
+/// A poster at a free mail provider (the judge's WEBMAIL): the stub floods' throwaway accounts.
+static WEBMAIL: LazyLock<Regex> = LazyLock::new(|| {
+    re(r"(?i)@(gmail|googlemail|yahoo|hotmail|outlook|live|msn|aol|gmx|web|t-online|freenet|mail|yandex|icloud|me|hushmail|protonmail|libero|orange|free|arcor|wanadoo|ymail|rocketmail|inbox|mail\.ru|bk|list)\.")
+});
+/// A name that reads as a release (the judge's RELEASE_LIKE): a year, a scene tag, a group suffix or dotted words.
+static RELEASE_LIKE: LazyLock<Regex> = LazyLock::new(|| {
+    re(r"(?i)(\b(19|20)[0-9][0-9]\b|[._-](720p|1080p|2160p|x264|x265|hevc|bluray|web-?dl|webrip|hdtv|xvid|s[0-9][0-9]e[0-9][0-9]|german|multi|crack|repack|proper|flac|mp3|xxx|mp4|retail|keygen|portable|incl|v[0-9]+(\.[0-9]+)+)\b|-[A-Za-z0-9]{2,}$|\.[A-Za-z0-9]+\.[A-Za-z0-9]+\.)")
+});
+
+/// RELEASE_LIKE with spaces as separators too: the tiny fakes are posted under names written either way.
+static RELEASE_LIKE_SPACED: LazyLock<Regex> = LazyLock::new(|| {
+    re(r"(?i)(\b(19|20)[0-9][0-9]\b|[._ -](720p|1080p|2160p|x264|x265|hevc|bluray|web-?dl|webrip|hdtv|xvid|s[0-9][0-9]e[0-9][0-9]|german|multi|crack|repack|proper|flac|mp3|xxx|mp4|retail|keygen|portable|incl|v[0-9]+(\.[0-9]+)+)\b|-[A-Za-z0-9]{2,}$|\.[A-Za-z0-9]+\.[A-Za-z0-9]+\.)")
+});
+
+/// The stub-archive rule (LOG.md (123)-(126), (128)): one unsplit archive, no par2, settled, posted to 2+ groups,
+/// under STUB_MAX_BYTES (the fakes stop at ~90 MB; the hash-named uploads, real albums and passworded videos above
+/// it don't), and its articles at most STUB_ARTICLE_BYTES with the release at least STUB_MIN_BYTES (Power-Post's
+/// 250 kB default; a one-article release always has small articles, so e-books and sidecars are exempt), or 4+
+/// groups, or a webmail poster.
+pub const STUB_ARTICLE_BYTES: u64 = 262_144;
+pub const STUB_MIN_BYTES: u64 = 1_000_000;
+pub const STUB_MAX_BYTES: u64 = 100_000_000;
+pub const STUB_WIDE_GROUPS: usize = 4;
+/// The nonsense-piece rule: one article holding one archive of NONSENSE_MIN_BYTES to NONSENSE_MAX_BYTES under a
+/// release-like name that isn't scrambled (the newznzb.* pieces; under 100 kB are cue/log sidecars and adverts).
+pub const NONSENSE_MIN_BYTES: u64 = 100_000;
+pub const NONSENSE_MAX_BYTES: u64 = 1_000_000;
+/// The tiny-archive rule's floor: a RAR holding one exe stub is 2 kB and up; below are empty archives and nfo rars.
+pub const TINY_MIN_BYTES: u64 = 2_000;
 
 pub struct JudgedFile<'a> {
     pub name: &'a str,
@@ -1245,6 +1542,79 @@ pub struct JudgedFile<'a> {
     pub groups: &'a [String],
     pub first_message_id: &'a str,
     pub poster: &'a str,
+    /// Articles seen and their bytes (the overview's :bytes).
+    pub segments: u32,
+    pub bytes: u64,
+}
+
+/// One unsplit archive and nothing else.
+fn single_archive<'a, 'b>(files: &'b [JudgedFile<'a>]) -> Option<&'b JudgedFile<'a>> {
+    match files {
+        [f] if test(&ARCHIVE, f.name) && !test(&VOLUME, f.name) => Some(f),
+        _ => None,
+    }
+}
+
+/// Groups whose communities post real single small archives (karaoke zips, e-books, comics, tabs, fonts) the way
+/// the stub floods do (Power-Post defaults, a crosspost or two): a release touching one is exempt from the
+/// stub-archive rule. Costs 16 of 32,308 known fakes in the 217k sample (LOG.md (127)).
+static STUB_EXEMPT_GROUPS: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)(sound|mp3|karaoke|music|audio|flac|lossless|e-?book|epub|comic|guitar|font|magazine)"));
+
+/// The legs of the stub-archive rule a settled single archive under STUB_MAX_BYTES trips ("A" article size, "X" 4+
+/// groups, "W" webmail), empty when none.
+pub fn stub_archive_legs(f: &JudgedFile) -> String {
+    let mut legs = String::new();
+    if f.groups.len() < 2 || f.bytes >= STUB_MAX_BYTES || f.groups.iter().any(|g| test(&STUB_EXEMPT_GROUPS, g)) {
+        return legs;
+    }
+    let article = f.bytes / u64::from(f.segments.max(1));
+    if article <= STUB_ARTICLE_BYTES && f.bytes >= STUB_MIN_BYTES {
+        legs.push('A');
+    }
+    if f.groups.len() >= STUB_WIDE_GROUPS {
+        legs.push('X');
+    }
+    if test(&WEBMAIL, f.poster) {
+        legs.push('W');
+    }
+    legs
+}
+
+/// An e-book by its name ("Author - Title (2020) (epub).rar"): one small article is a book, not a nonsense piece.
+static EBOOK: LazyLock<Regex> = LazyLock::new(|| re(r"(?i)\b(epub|mobi|azw3?|pdf|djvu|cbr|cbz|e-?books?)\b"));
+
+/// A nonsense piece: one article, one small archive under a release-like, unscrambled name without spaces (the
+/// newznzb.* pieces are scene-styled or random with a "-XX" suffix; books and albums are written with spaces).
+pub fn is_nonsense_piece(f: &JudgedFile) -> bool {
+    f.total_parts == 1
+        && f.segments == 1
+        && (NONSENSE_MIN_BYTES..NONSENSE_MAX_BYTES).contains(&f.bytes)
+        && !f.name.contains(' ')
+        && test(&RELEASE_LIKE, &file_stem(f.name))
+        && !test(&EBOOK, f.name)
+        && !looks_scrambled(f.name)
+        && !is_random_file(f.name)
+}
+
+/// Words of the real tiny archives under release-like names: subtitles, cracks, trainers and keygens (a few kB of
+/// exe is the content), covers and sidecars.
+static TINY_CONTENT: LazyLock<Regex> = LazyLock::new(|| {
+    re(r"(?i)\b(subs?|subtitles?|subpack|srt|untertitel|ondertitels?|ondertiteling|legendas?|napisy|sottotitoli|cracks?|cracked|keygen|keymaker|trainer|patch|patched|serials?|activator|loader|nocd|no-cd|fix|covers?|artwork|booklet|scans?|proof|sample|info|readme)\b")
+});
+
+/// A tiny archive: one article of TINY_MIN_BYTES to NONSENSE_MIN_BYTES holding one archive under a release-like,
+/// unscrambled name (spaces allowed), outside the groups that post real small archives, and without an e-book,
+/// subtitle, crack or sidecar word: the 2015-17 exe-stub flood in miniature (LOG.md (131)).
+pub fn is_tiny_archive(f: &JudgedFile) -> bool {
+    f.total_parts == 1
+        && f.segments == 1
+        && (TINY_MIN_BYTES..NONSENSE_MIN_BYTES).contains(&f.bytes)
+        && !f.groups.iter().any(|g| test(&STUB_EXEMPT_GROUPS, g))
+        && test(&RELEASE_LIKE_SPACED, &file_stem(f.name))
+        && !test(&EBOOK, f.name)
+        && !test(&TINY_CONTENT, f.name)
+        && !looks_scrambled(f.name)
+        && !is_random_file(f.name)
 }
 
 pub enum Verdict {
@@ -1260,12 +1630,19 @@ pub struct Rules {
     pub priorities: rustc_hash::FxHashMap<String, u32>,
     /// "tool\tgroup"
     pub dropped_signatures: rustc_hash::FxHashSet<String>,
+    /// "tool\tgroup\tklass" (pending_class): unnamed releases of these classes don't go to the pending store.
+    pub dropped_classes: rustc_hash::FxHashSet<String>,
     pub dropped_tools: rustc_hash::FxHashSet<String>,
     pub per_article_max_parts: u32,
     pub sample_percent: f64,
     pub drop_executables: bool,
     /// Fragments of the poster lowercased without punctuation (droppedPosters).
     pub dropped_posters: Vec<String>,
+    /// The stub-archive rule (dropStubArchives), the nonsense-piece rule (dropNonsensePieces) and the tiny-archive
+    /// rule (dropTinyArchives).
+    pub drop_stub_archives: bool,
+    pub drop_nonsense_pieces: bool,
+    pub drop_tiny_archives: bool,
 }
 
 impl Rules {
@@ -1275,11 +1652,15 @@ impl Rules {
             version: 1,
             priorities,
             dropped_signatures: Default::default(),
+            dropped_classes: Default::default(),
             dropped_tools: ["ngPost", "hex32", "yuun", "JBinUp", "random"].iter().map(|t| t.to_string()).collect(),
             per_article_max_parts: 3,
             sample_percent: 1.0,
             drop_executables: true,
             dropped_posters: ["celebmovie", "celebwow", "celebfx", "iluvcelebs", "nudecelebs", "celebsextapes"].iter().map(|p| p.to_string()).collect(),
+            drop_stub_archives: true,
+            drop_nonsense_pieces: true,
+            drop_tiny_archives: true,
         }
     }
 }
@@ -1289,8 +1670,9 @@ pub fn poster_words(poster: &str) -> String {
     poster.chars().filter(|c| c.is_ascii_alphanumeric()).map(|c| c.to_ascii_lowercase()).collect()
 }
 
-/// RuleSet.judge.
-pub fn judge(release_key: &str, files: &[JudgedFile], rules: &Rules) -> Verdict {
+/// RuleSet.judge. `settled`: the window closed on the release (a par2 can still join an open one), which the
+/// single-archive rules need.
+pub fn judge(release_key: &str, files: &[JudgedFile], settled: bool, rules: &Rules) -> Verdict {
     let sample = explored(release_key, rules.sample_percent);
     if let (false, Some(first)) = (rules.dropped_posters.is_empty(), files.first()) {
         let words = poster_words(first.poster);
@@ -1304,6 +1686,20 @@ pub fn judge(release_key: &str, files: &[JudgedFile], rules: &Rules) -> Verdict
         let key = std::str::from_utf8(&c[1]).unwrap().to_lowercase();
         return Verdict::Drop { rule: "executable", key, sample };
     }
+    if let (true, Some(f)) = (settled, single_archive(files)) {
+        if rules.drop_stub_archives {
+            let legs = stub_archive_legs(f);
+            if !legs.is_empty() {
+                return Verdict::Drop { rule: "stub-archive", key: legs, sample };
+            }
+        }
+        if rules.drop_nonsense_pieces && is_nonsense_piece(f) {
+            return Verdict::Drop { rule: "nonsense-piece", key: upload_tool(f.first_message_id).to_string(), sample };
+        }
+        if rules.drop_tiny_archives && is_tiny_archive(f) {
+            return Verdict::Drop { rule: "tiny-archive", key: upload_tool(f.first_message_id).to_string(), sample };
+        }
+    }
     if files.iter().any(|f| !is_random_file(f.name)) {
         return Verdict::Keep;
     }
@@ -1334,6 +1730,29 @@ mod tests {
         assert_eq!(file_stem_re("Morning Lovers.wmv.058"), "Morning Lovers");
     }
 
+    #[test]
+    fn scrambled_names() {
+        for n in ["nYDeXeF49dieZ1J7flK-JFzZ3j.ZEDsr", "bslXlwgPiqZadhRa", "PNSVWSV-xLaNv6o5qS840e3lwSZyH", "Biq-YsXu8G33ktt"] {
+            assert!(looks_scrambled(n), "{n}");
+        }
+        // Short ones of the spam too (8% of alt.binaries.storage's): too few switches to tell.
+        for n in ["A1Monitor v6.2.2", "PDF2Word v1.3", "Some.Show.S01E01.1080p.WEB.part1.rar", "MacGyver.S01E01.mkv", "Jimmy and the Crawler (mobi)", "Q25m.PDUYtbA", "82WH.2iuTK"] {
+            assert!(!looks_scrambled(n), "{n}");
+        }
+    }
+
+    #[test]
+    fn display_names() {
+        let porn4you = "webmaster@hdbangvidz.com (Porn4You)";
+        let a = release_identity(porn4you, "Babe with Plump Bubble Butt Fucked and Facialized 002/477] - 0178.jpg", "0178.jpg");
+        let b = release_identity(porn4you, "Babe with Plump Bubble Butt Fucked and Facialized 003/477] - 0179.jpg", "0179.jpg");
+        assert_eq!(a.name, "Babe with Plump Bubble Butt Fucked and Facialized");
+        assert_eq!(a.key, b.key);
+        assert_eq!(display_name("[Group] Show (2020)"), "[Group] Show (2020)");
+        assert_eq!(display_name("( Show [x] name"), "Show [x] name");
+        assert_eq!(display_name("]"), "]");
+    }
+
     #[test]
     fn remainders() {
         assert_eq!(subject_remainder(r#"Some.Release [01/45] - "x.rar" yEnc (1/50) 7168"#), "Some Release");
@@ -1446,6 +1865,46 @@ mod tests {
         assert_eq!(id(r#"[2/6] - "x.jpg" yEnc"#, "x.jpg").key, id(r#"[1/6] - "x.part01.rar" yEnc"#, "x.part01.rar").key);
     }
 
+    #[test]
+    fn pending_classes() {
+        use super::pending_class;
+        // Size bucket × shape × partial: a 5 MB hex name whole, the same partial, a 500 MB scramble with separators.
+        assert_eq!(pending_class("c04212a389cf69edc54767355ab5285f", 5_000_000, false), 15);
+        assert_eq!(pending_class("c04212a389cf69edc54767355ab5285f", 5_000_000, true), 16);
+        assert_eq!(pending_class("Zx5AjNaK-mh7CFg6Hf8uu", 500_000_000, false), 3 * 14 + 5 * 2 + 1);
+        assert_eq!(pending_class("1650777167", 100, false), 3);
+        assert_eq!(pending_class("mrrtmzrzmi4dgmrzmyytqmlbztczbx", 2_000_000_000, false), 4 * 14 + 2 * 2 + 1);
+        assert_eq!(pending_class("PsNdnYiugfZ.rar", 20_000_000, false), 2 * 14 + 3 * 2 + 1);
+        assert_eq!(pending_class("n5d1ueIf34X", 20_000_000, true), 2 * 14 + 4 * 2 + 2);
+        assert_eq!(pending_class("sgkra vU7u2yuw9LDLCjDu04pOez", 999, false), 13);
+    }
+
+    #[test]
+    fn unreadable_names() {
+        use super::is_unreadable_name;
+        // Scrambles with separators, spaces, a random "extension", a packaging word, a hex with an addon (LOG.md (129)).
+        for n in [
+            "Zx5AjNaK-mh7CFg6Hf8uu.part01.rar", "yWjfnT2ujG4wJj3fG46o0FLzsgkvhmrTsAfW2.E0WsOU", "PsNdnYiugfZ.rar", "phfplC9y07U2yrj", "n5d1ueIf34X",
+            "tJqr1tq7xK", "Description - Lt5nHjlo3RiuBb3C.par2", "bd5b4a59f3aa87114dcf75c0-addon.rar", "mrrtmzrzmi4dgmrzmyytqmlbztczbx",
+            "AA-NaR16302830434zxP32108309f3Ab", "_QBs01_MqZYnYwM_iAyRquhc-", "sgkra vU7u2yuw9LDLCjDu04pOez", "Jk5Pud.zQdNj", "7uKsD_NRgT-G_Uc_p8",
+            "NtpUgCmS_Xa1B2Pk", "{E11BF8E9-E2C9-4DAE-A1B2-8027BC465F3F}", "IkYovjrtZrItjQ.mkv", "ff75ce3f16c746c18b2d78cf.6c52f",
+        ] {
+            assert!(is_unreadable_name(n), "{n}");
+        }
+        // Real names: scene tags, CamelCase, codes, leet, an acronym, a hash beside words.
+        for n in [
+            "Dark.Matter.2024.S02E10.No.Quiet.Life.1080p.ATVP.WEB-DL.DDP5.1.Atmos.H.264-Kitsune", "Paul McCartney - Band on the Run",
+            "ShesSoWild - Ester Ana 1080p", "MyDadsHotGirlfriend.13.02.01.Luna.Star.XXX.1080p.x264-SEXORS", "cFosSpeed 11.0 Build 2420 x86 x64 incl Patch.rar",
+            "Rocketship-Hurt_Feelings-WEB-2026-PreMaxXxing", "Drum-and-BassDJ_Samurai-Sound_With_Fire_Artemis-(FQY026)-WEB-2006-PRESCRiPTiON",
+            "Kleverig-H4rd-N12uk3b18-luc16037-720p-smpwr", "Eclipse_IDE_2023-12_R_win32_x86_64.zip", "PartsHPGL 1.06c", "Go4Customer Fax Machine v1.2b",
+            "AdBeGone 1.2", "3DMark03 320", "COD4MW2REMASTERED", "Movie.2020.1080p-GRP [a3f9e2b1]", "EPL__NZBSPLIT__f1fdb433772a1a7771d6b73c875009ff__NZBSPLIT__EPL.2026.Liverpool.vs.Brighton.1080pEN50fps",
+            "iPhone", "Some.Show.S01E01.1080p.mkv", "FastFilms 2.04", "BestCrypt NP v4.0", "EasyHTML 2.35 German", "SchiffsModell - September 2026",
+            "CrossStitcher - Issue 440, October 2026", "Laura15Juli2026", "IHaveAWife - Isa Bella", "DickHDDaily - Santana Red", "1904-SmithsWork-Hasluck",
+        ] {
+            assert!(!is_unreadable_name(n), "{n}");
+        }
+    }
+
     #[test]
     fn random_names() {
         assert!(looks_random("prvjr7hapd0i1ny14pz4"));
@@ -1486,8 +1945,8 @@ mod poster_rule_tests {
 
     fn verdict(poster: &str, rules: &Rules) -> Option<(&'static str, String)> {
         let groups = vec!["alt.binaries.erotica".to_string()];
-        let files = [JudgedFile { name: "!index0001.jpg", total_parts: 1, groups: &groups, first_message_id: "<a@b>", poster }];
-        match judge("k", &files, rules) {
+        let files = [JudgedFile { name: "!index0001.jpg", total_parts: 1, groups: &groups, first_message_id: "<a@b>", poster, segments: 1, bytes: 200_000 }];
+        match judge("k", &files, true, rules) {
             Verdict::Keep => None,
             Verdict::Drop { rule, key, .. } => Some((rule, key)),
         }
@@ -1507,6 +1966,111 @@ mod poster_rule_tests {
     }
 }
 
+#[cfg(test)]
+mod single_archive_rule_tests {
+    use super::{judge, JudgedFile, Rules, Verdict};
+
+    const NYUU: &str = "FmFwIaTpGgUiSfYcMwFlMeRo-1790637260478@nyuu";
+
+    fn verdict(name: &str, poster: &str, groups: usize, segments: u32, bytes: u64, settled: bool, rules: &Rules) -> Option<(&'static str, String)> {
+        let groups: Vec<String> = (0..groups).map(|i| format!("alt.binaries.g{i}")).collect();
+        let files = [JudgedFile { name, total_parts: segments, groups: &groups, first_message_id: NYUU, poster, segments, bytes }];
+        match judge("k", &files, settled, rules) {
+            Verdict::Keep => None,
+            Verdict::Drop { rule, key, .. } => Some((rule, key)),
+        }
+    }
+
+    #[test]
+    fn stub_archives_by_article_size_groups_or_webmail() {
+        let rules = Rules::defaults(Default::default());
+        // Power-Post's 250 kB articles, two groups, 12 MB: the 2016 wave.
+        assert_eq!(verdict("Chat Watch 1.2.1.zip", "jahnunsinos007@gmail.com (Jahnunsinos007)", 2, 48, 12_000_000, true, &rules), Some(("stub-archive", "AW".to_string())));
+        assert_eq!(verdict("Some.Movie.2014.720p.BluRay-GRP.rar", "Yenc@power-post.org (Yenc-PP-A&A)", 2, 48, 12_000_000, true, &rules), Some(("stub-archive", "A".to_string())));
+        // Five groups, 319 kB in two articles: the 2015 a.b.cracks wave (under 1 MB, so the groups leg alone).
+        assert_eq!(verdict("Big Hero 6 (2014) 720p BDRip [COMPRESSED].rar", "Yenc@power-post.org (Yenc-PP-A&A)", 5, 2, 319_000, true, &rules), Some(("stub-archive", "X".to_string())));
+        // A webmail poster with 650 kB articles: the 2014 "leaked Sex Video" wave.
+        assert_eq!(verdict("Leaked Sex Video.rar", "first.last88@gmail.com", 2, 400, 260_000_000, true, &rules), None, "over 100 MB without the article-size leg");
+        assert_eq!(verdict("Leaked Sex Video.rar", "first.last88@gmail.com", 2, 100, 65_000_000, true, &rules), Some(("stub-archive", "W".to_string())));
+        // Settled only: a par2 can still join an open release.
+        assert_eq!(verdict("Chat Watch 1.2.1.zip", "jahnunsinos007@gmail.com", 2, 48, 12_000_000, false, &rules), None);
+    }
+
+    #[test]
+    fn small_and_single_group_and_split_archives_are_kept() {
+        let rules = Rules::defaults(Default::default());
+        // An e-book: one or two small articles always have small articles (LOG.md (126)).
+        assert_eq!(verdict("Maeve Binchy - Tara Road (epub).rar", "kie@home.now (KiE)", 2, 2, 521_000, true, &rules), None);
+        // One group: most real single rars.
+        assert_eq!(verdict("Chat Watch 1.2.1.zip", "jahnunsinos007@gmail.com", 1, 48, 12_000_000, true, &rules), None);
+        // Modern 700 kB articles, no webmail, two groups.
+        assert_eq!(verdict("Some.Movie.2014.720p.BluRay-GRP.rar", "poster@example.org", 2, 100, 70_000_000, true, &rules), None);
+        // A volume, not a single archive; a hash-named 2 GB upload from a gmail-looking poster.
+        assert_eq!(verdict("Some.Movie.2014.part01.rar", "a@gmail.com", 2, 48, 12_000_000, true, &rules), None);
+        assert_eq!(verdict("cbcd32359460df1fa9b7d01d8720.rar", "b648de0d1c0f7a9a77be273d@gmail.com", 2, 3163, 2_340_000_000, true, &rules), None);
+        // ROT13-named volumes of passworded videos in a.b.erotica, 258 kB articles, two groups, 1-270 MB; and the
+        // 100 MB cap on every leg (LOG.md (128)).
+        assert_eq!(verdict("Ghful.17.12.07.Qnvvfl.Fgbar.cneg3.rar", "nospam@nospam.nl (EvH)", 2, 5, 1_147_000, true, &rules), None);
+        assert_eq!(verdict("Ghful.16.07.05.Wbwb.Xvff.cneg1.rar", "nospam@nospam.nl (EvH)", 2, 1051, 271_000_000, true, &rules), None);
+        assert_eq!(verdict("Tinasoft EasyCafe 2.2.14 by OmerSalaj.rar", "Seven7i@mail.com ({Seven7i})", 4, 463, 117_000_000, true, &rules), None);
+        assert_eq!(verdict("Tipard DVD Ripper - 10 0 12 x86 x64 incl Patch.rar", "Seven7i@mail.com ({Seven7i})", 4, 244, 62_000_000, true, &rules), Some(("stub-archive", "AXW".to_string())));
+        let off = Rules { drop_stub_archives: false, ..Rules::defaults(Default::default()) };
+        assert_eq!(verdict("Chat Watch 1.2.1.zip", "jahnunsinos007@gmail.com", 2, 48, 12_000_000, true, &off), None);
+        // A karaoke zip crossposted to a.b.test with Power-Post's articles: its group exempts it (LOG.md (127)).
+        let groups = vec!["alt.binaries.sounds.karaoke".to_string(), "alt.binaries.test".to_string()];
+        let files = [JudgedFile { name: "KV-1055806 - Walker Hayes - U Gurl.zip", total_parts: 13, groups: &groups, first_message_id: NYUU, poster: "home@home.com (TexasDj)", segments: 13, bytes: 3_300_000 }];
+        assert!(matches!(judge("k", &files, true, &rules), Verdict::Keep));
+    }
+
+    #[test]
+    fn nonsense_pieces() {
+        let rules = Rules::defaults(Default::default());
+        // One 790 kB article under a release-like name: the newznzb.* pieces.
+        assert_eq!(verdict("Some.Show.S01E01.1080p.WEB-DL.rar", "poster@example.org", 1, 1, 790_000, true, &rules), Some(("nonsense-piece", "nyuu".to_string())));
+        // A 3 kB cue/log sidecar rar with a year in its name: under this rule's floor (the tiny-archive rule's outside the music groups).
+        assert_eq!(verdict("The Smiths-the queen is dead(Remaster)(2011).rar", "Yenc@power-post.org (Sir Cedrik)", 1, 1, 3_000, true, &rules), Some(("tiny-archive", "nyuu".to_string())));
+        // A book (a year or a format in its name, written with spaces), a dotted one too: not a piece.
+        assert_eq!(verdict("Kevin Ashman - [Roman 02] - The Rise of Caratacus (2020) (epub).rar", "fake@address.com (Psychopath)", 2, 1, 303_000, true, &rules), None);
+        assert_eq!(verdict("One.Wrong.Turn.Deanna.Lynn.Sletten.epub.rar", "fake@address.com (Psychopath)", 1, 1, 382_000, true, &rules), None);
+        assert_eq!(verdict("Vieweg.Teubner.Verlag.-.Numerische.Mathematik.2011.German.RETAIL.eBook-BitBook.rar", "saggir@domain.com", 1, 1, 113_000, true, &rules), None);
+        // Two articles, or no release-like name, or scrambled: not this rule.
+        assert_eq!(verdict("Some.Show.S01E01.1080p.WEB-DL.rar", "poster@example.org", 1, 2, 790_000, true, &rules), None);
+        assert_eq!(verdict("holiday snaps.rar", "poster@example.org", 1, 1, 790_000, true, &rules), None);
+        assert_eq!(verdict("nYDeXeF49dieZ1J7flK-JFzZ3j.rar", "poster@example.org", 1, 1, 790_000, true, &rules), None);
+        let off = Rules { drop_nonsense_pieces: false, ..Rules::defaults(Default::default()) };
+        assert_eq!(verdict("Some.Show.S01E01.1080p.WEB-DL.rar", "poster@example.org", 1, 1, 790_000, true, &off), None);
+    }
+
+    #[test]
+    fn tiny_archives() {
+        let rules = Rules::defaults(Default::default());
+        // One article of 2-100 kB under a release name, dotted or with spaces: a RAR holding "<title>.exe" (LOG.md (131)).
+        assert_eq!(verdict("Humans.S01E04.German.DD20.Dubbed.DL.1080p.BD.x264-TVS.rar", "Yanc@power-post.org <Yanc@power-post.org>", 1, 1, 15_084, true, &rules), Some(("tiny-archive", "nyuu".to_string())));
+        assert_eq!(verdict("Cicero Magazin fuer politische Kultur Mai 05 2016.rar", "WarezGeek@WarezGeek.org (WarezGeek)", 1, 1, 17_478, true, &rules), Some(("tiny-archive", "nyuu".to_string())));
+        // A scene name without a year, a tag or a group suffix: not release-like, so kept (the flood's porn names mostly carry one).
+        assert_eq!(verdict("[BigTitsAtWork Brazzers] Eva Angelina (Camera Cums In Handy).rar", "Yanc@power-post.org", 2, 1, 23_449, true, &rules), None);
+        assert_eq!(verdict("Graphisoft.Archicad.v14.0.GERMAN.iSO.X86-PillePalle.rar", "WarezGeek@WarezGeek.org (WarezGeek)", 1, 1, 7_869, true, &rules), Some(("tiny-archive", "nyuu".to_string())));
+        // Subtitles, a crack, a trainer, a cover, an e-book: the real tiny archives.
+        assert_eq!(verdict("The.Big.Bang.Theory.S04E13.720p.HDTV.subs.rar", "henk@example.org", 1, 1, 38_000, true, &rules), None);
+        assert_eq!(verdict("Tipard.DVD.Ripper.v10.0.12.Crack-XXX.rar", "poster@example.org", 1, 1, 60_000, true, &rules), None);
+        assert_eq!(verdict("Far.Cry.4.v1.7.0.Plus.12.Trainer-FLiNG.rar", "poster@example.org", 1, 1, 45_000, true, &rules), None);
+        assert_eq!(verdict("Some Band - Album (2011) covers.rar", "poster@example.org", 1, 1, 45_000, true, &rules), None);
+        assert_eq!(verdict("One.Wrong.Turn.Deanna.Lynn.Sletten.epub.rar", "fake@address.com (Psychopath)", 1, 1, 38_000, true, &rules), None);
+        // A karaoke zip in its group; a plain name; a scrambled one; two articles; under the floor; above 100 kB is the nonsense-piece rule's.
+        let groups = vec!["alt.binaries.sounds.karaoke".to_string()];
+        let files = [JudgedFile { name: "KV-1055806 - Walker Hayes - U Gurl (2017).zip", total_parts: 1, groups: &groups, first_message_id: NYUU, poster: "home@home.com", segments: 1, bytes: 60_000 }];
+        assert!(matches!(judge("k", &files, true, &rules), Verdict::Keep));
+        assert_eq!(verdict("holiday snaps.rar", "poster@example.org", 1, 1, 20_000, true, &rules), None);
+        assert_eq!(verdict("nYDeXeF49dieZ1J7flK-JFzZ3j.rar", "poster@example.org", 1, 1, 20_000, true, &rules), None);
+        assert_eq!(verdict("Humans.S01E04.German.DL.1080p.BD.x264-TVS.rar", "poster@example.org", 1, 2, 20_000, true, &rules), None);
+        assert_eq!(verdict("Humans.S01E04.German.DL.1080p.BD.x264-TVS.rar", "poster@example.org", 1, 1, 1_500, true, &rules), None);
+        assert_eq!(verdict("Humans.S01E04.German.DL.1080p.BD.x264-TVS.rar", "poster@example.org", 1, 1, 150_000, true, &rules), Some(("nonsense-piece", "nyuu".to_string())));
+        assert_eq!(verdict("Humans.S01E04.German.DL.1080p.BD.x264-TVS.rar", "poster@example.org", 1, 1, 20_000, false, &rules), None);
+        let off = Rules { drop_tiny_archives: false, ..Rules::defaults(Default::default()) };
+        assert_eq!(verdict("Humans.S01E04.German.DL.1080p.BD.x264-TVS.rar", "poster@example.org", 1, 1, 20_000, true, &off), None);
+    }
+}
+
 #[cfg(test)]
 mod image_set_tests {
     use super::{release_identity, sized_name, subject_remainder, subject_remainder_re};

edge/src/wire.rs

@@ -1,7 +1,7 @@
 //! A batch as the coordinator takes it (lib/edge/wire.ts toWire): JSON with
 //! the TypeScript names, chunk payloads in base64.
 
-use crate::indexer::{Batch, Candidate, Fragment, ReleaseChunk};
+use crate::indexer::{Batch, Candidate, ReleaseChunk};
 use base64::Engine;
 use serde::Serialize;
 
@@ -24,6 +24,11 @@ struct WireRelease<'a> {
     last_article: i64,
     readable: bool,
     has_par2: bool,
+    finished: bool,
+    edge: bool,
+    partial: bool,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    tool: Option<&'a str>,
     payload: String,
     #[serde(skip_serializing_if = "Option::is_none")]
     rule: Option<&'a str>,
@@ -39,6 +44,7 @@ struct WireCandidate<'a> {
     kind: &'a str,
     message_id: &'a str,
     file: &'a str,
+    bytes: i64,
 }
 
 #[derive(Serialize)]
@@ -49,30 +55,6 @@ struct WireCount<'a> {
     files: u64,
 }
 
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct WireFragmentRelease<'a> {
-    key: &'a str,
-    name: &'a str,
-    source: &'a str,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    file_total: Option<i64>,
-}
-
-#[derive(Serialize)]
-#[serde(rename_all = "camelCase")]
-struct WireFragment<'a> {
-    file_key: &'a str,
-    name: &'a str,
-    poster: &'a str,
-    subject: &'a str,
-    total_parts: u32,
-    groups: &'a [String],
-    release: Option<WireFragmentRelease<'a>>,
-    posted_at: Option<i64>,
-    segments: Vec<(u32, &'a str, i64, i64)>,
-}
-
 #[derive(Serialize)]
 #[serde(rename_all = "camelCase")]
 struct WireBatch<'a> {
@@ -85,7 +67,6 @@ struct WireBatch<'a> {
     samples: Vec<WireRelease<'a>>,
     candidates: Vec<WireCandidate<'a>>,
     counts: Vec<WireCount<'a>>,
-    fragments: Vec<WireFragment<'a>>,
 }
 
 fn release<'a>(r: &'a ReleaseChunk, rule: Option<(&'a str, &'a str)>) -> WireRelease<'a> {
@@ -106,6 +87,10 @@ fn release<'a>(r: &'a ReleaseChunk, rule: Option<(&'a str, &'a str)>) -> WireRel
         last_article: r.last_article,
         readable: r.readable,
         has_par2: r.has_par2,
+        finished: r.finished,
+        edge: r.edge,
+        partial: r.partial,
+        tool: r.tool,
         payload: base64::engine::general_purpose::STANDARD.encode(&r.payload),
         rule: rule.map(|(rule, _)| rule),
         rule_key: rule.map(|(_, key)| key),
@@ -113,22 +98,7 @@ fn release<'a>(r: &'a ReleaseChunk, rule: Option<(&'a str, &'a str)>) -> WireRel
 }
 
 fn candidate(c: &Candidate) -> WireCandidate<'_> {
-    WireCandidate { release_key: &c.release_key, upload: c.upload, kind: c.kind, message_id: &c.message_id, file: &c.file }
-}
-
-fn fragment(f: &Fragment) -> WireFragment<'_> {
-    WireFragment {
-        file_key: &f.file_key,
-        name: &f.name,
-        poster: &f.poster,
-        subject: &f.subject,
-        total_parts: f.total_parts,
-        groups: &f.groups,
-        // A named file always has a release; fileTotal only when truthy, as fragmentOf.
-        release: Some(WireFragmentRelease { key: &f.release.key, name: &f.release.name, source: f.release.source, file_total: f.release.file_total_truthy() }),
-        posted_at: f.posted_at,
-        segments: f.segments.iter().map(|(n, id, bytes, article)| (*n, id.as_str(), *bytes, *article)).collect(),
-    }
+    WireCandidate { release_key: &c.release_key, upload: c.upload, kind: c.kind, message_id: &c.message_id, file: &c.file, bytes: c.bytes }
 }
 
 /// toWire(batch) as JSON bytes.
@@ -143,7 +113,6 @@ pub fn to_json(batch: &Batch) -> Vec<u8> {
         samples: batch.samples.iter().map(|(r, rule, key)| release(r, Some((rule, key.as_str())))).collect(),
         candidates: batch.candidates.iter().map(candidate).collect(),
         counts: batch.counts.iter().map(|(rule, key, articles, files)| WireCount { rule, key, articles: *articles, files: *files }).collect(),
-        fragments: batch.fragments.iter().map(fragment).collect(),
     };
     serde_json::to_vec(&wire).unwrap()
 }

lib/edge/index.ts

@@ -1,8 +1,7 @@
 export { RangeIndexer, postedSpan, chunkOf, candidatesOf, releaseKeyOf, uploadsOf, UPLOAD_GAP_MS, type EdgeBatch, type ReleaseChunk, type DroppedChunk, type ProbeCandidate, type CandidateKind, type RuleCount, type Fragment, type RangeIndexerOptions } from './range_indexer.js'
 export { encodeFrame, decodeFrames, isKeepalive, KEEPALIVE, type Frame } from './stream.js'
 export { toWire, fromWire, type WireBatch } from './wire.js'
-export type { Lease } from './lease.js'
-export { mergeFragments } from './merge.js'
+export type { Lease, GroupLease, Lane, NextRange } from './lease.js'
 export type { EdgeSegment, FileMeta } from './scan.js'
 export { throttle } from './throttle.js'
 export { filterOverviews, undelta, type ChunkDrops } from './corpus_filter.js'

lib/edge/lease.ts

@@ -5,3 +5,27 @@ export interface Lease {
   range: { id: number; group: string; start: number; end: number; next: number }
   rules: Rules
 }
+
+/**
+ * A group's two cursors (REBUILD.md, "Voyager 3: window indexing", 1): `history` works the ranges the lease bounds
+ * hand out (the windows, the history), `tail` the tail planner's ranges. Each lane has its own holder, window and
+ * snapshot; the group routes take `?lane=` (history by default).
+ */
+export type Lane = 'history' | 'tail'
+
+/**
+ * A group lease (REBUILD.md, "Voyager 3: window indexing", 1): the group a worker thread holds, the lane, the rules,
+ * and the snapshot of the lane's window when there is one (its next article; the body comes from GET
+ * groups/:id/snapshot).
+ */
+export interface GroupLease {
+  group: { id: number; name: string; priority: number }
+  lane: Lane
+  rules: Rules
+  snapshot: { next: number } | null
+}
+
+/** The group's next range in article order, leased to the holder (POST groups/:id/next; 204 when caught up). */
+export interface NextRange {
+  range: { id: number; start: number; end: number; next: number }
+}

lib/edge/merge.ts

@@ -1,35 +0,0 @@
-import type { AggregatedFile } from '../files/index.js'
-import type { Fragment } from './range_indexer.js'
-import type { EdgeSegment, FileMeta } from './scan.js'
-
-/**
- * The fragments of one file from neighbouring ranges, put together
- * (REBUILD.md, fragments): the complete file when every part from 1 to its
- * total is there (the first article seen of a part wins), else null.
- */
-export function mergeFragments(fragments: Fragment[]): AggregatedFile<EdgeSegment> | null {
-  if (fragments.length === 0) return null
-  const first = fragments[0]
-  const meta: FileMeta = {
-    fileKey: first.fileKey,
-    name: first.name,
-    poster: first.poster,
-    subject: first.subject,
-    totalParts: first.totalParts,
-    groups: first.groups,
-    release: first.release ? { ...first.release, source: first.release.source as any } : null,
-  }
-  const parts = new Map<number, EdgeSegment>()
-  for (const fragment of fragments) {
-    const postedAt = fragment.postedAt ?? Number.NaN
-    for (const [number, messageId, bytes, articleNumber] of fragment.segments) {
-      const known = parts.get(number)
-      if (!known || articleNumber < known.articleNumber) {
-        parts.set(number, { fileKey: meta.fileKey, number, totalParts: meta.totalParts, messageId, bytes, articleNumber, postedAt, file: meta })
-      }
-    }
-  }
-  if (parts.size < meta.totalParts) return null
-  for (let n = 1; n <= meta.totalParts; n++) if (!parts.has(n)) return null
-  return { fileKey: meta.fileKey, segments: [...parts.values()].sort((a, b) => a.number - b.number) }
-}

lib/edge/range_indexer.ts

@@ -1,7 +1,9 @@
 import { hash } from 'node:crypto'
 import type { Overview } from '../nntp/index.js'
 import { FileAggregator, type AggregatedFile, isArchiveStart, isLaterVolume, sevenZipVolume } from '../files/index.js'
-import { RuleSet, isRandomFile, uploadTool, type Rules } from '../rules/index.js'
+import { RuleSet, isRandomFile, isScrambledFile, uploadTool, type Rules } from '../rules/index.js'
+import { classKey, isUnreadableName, pendingClass } from '../release/index.js'
+import { explored } from '../rules/upload_tools.js'
 import { SegmentScanner, scanOverviewText, type EdgeSegment } from './scan.js'
 import { encodeChunk, type ChunkFile } from '../chunks/index.js'
 
@@ -26,6 +28,20 @@ export interface ReleaseChunk {
   /** A file with a readable name; a par2 file. */
   readable: boolean
   hasPar2: boolean
+  /**
+   * Voyager 3: the window closed on the release (default), or it is a tail release sent provisionally, finished by a
+   * later chunk of the same release key (the totals then are the release's; the payload only the files not sent yet).
+   */
+  finished?: boolean
+  /** Still open at its block's end, or posted within the window of its start: the block merger joins it with the neighbour block's part. */
+  edge?: boolean
+  /** A file of the release misses parts (the window closed on it): not complete, whatever the counter says. */
+  partial?: boolean
+  /**
+   * Voyager 3, an unnamed release (readable false): the upload tool from its first article's message-id, its
+   * signature with the group; the payload is empty (no chunk until a probe names it: pending_releases).
+   */
+  tool?: string
   payload: Buffer
 }
 
@@ -44,11 +60,13 @@ export interface ProbeCandidate {
   kind: CandidateKind
   messageId: string
   file: string
+  /** The target article's bytes: the cheapest probe first within a class (LOG.md (130)). */
+  bytes?: number
 }
 
 /** What a rule dropped in a batch (REBUILD.md, rules and recovery). */
 export interface RuleCount {
-  rule: 'crosspost' | 'obfuscated' | 'poster' | 'executable' | 'garbage' | 'per-article' | 'incomplete'
+  rule: 'crosspost' | 'obfuscated' | 'poster' | 'executable' | 'garbage' | 'per-article' | 'stub-archive' | 'nonsense-piece' | 'tiny-archive' | 'incomplete'
   key: string
   articles: number
   files: number
@@ -88,7 +106,8 @@ export interface EdgeBatch {
   samples: DroppedChunk[]
   candidates: ProbeCandidate[]
   counts: RuleCount[]
-  fragments: Fragment[]
+  /** The range path's cut-off files (the TypeScript indexer); the Rust edge's window has none, the store ignores them. */
+  fragments?: Fragment[]
 }
 
 export interface RangeIndexerOptions {
@@ -108,6 +127,11 @@ export interface RangeIndexerOptions {
    * chunks seen so far), as uploads in busy groups spread over minutes.
    */
   edgeArticles?: number
+  /**
+   * A file given up short of parts is a file of its release all the same (partial), as the Rust edge's window writes
+   * it, not a fragment: the pending namer reads such a release again (LOG.md 2026-10-29 (121)).
+   */
+  partialFiles?: boolean
 }
 
 const MAX_CANDIDATES = 4
@@ -203,7 +227,7 @@ export class RangeIndexer {
 
   /** Files completed and fragments cut since the last take: how big the next batch is. */
   get pending(): number {
-    let files = this.batch.fragments.length
+    let files = this.batch.fragments!.length
     for (const list of this.completed.values()) files += list.length
     return files
   }
@@ -226,12 +250,34 @@ export class RangeIndexer {
     this.completed = new Map()
     for (const [key, all] of byRelease) {
       uploadsOf(all).forEach((list, upload) => {
+        // Settled as far as this indexer knows: it judges a release's files as each batch completes them (the Rust
+        // window passes whether the window closed on the release and no block edge cut it).
         const verdict = this.rules.judge(
           key,
-          list.map((f) => ({ name: f.segments[0].file.name, totalParts: f.segments[0].totalParts, groups: f.segments[0].file.groups, firstMessageId: f.segments[0].messageId, poster: f.segments[0].file.poster }))
+          list.map((f) => ({
+            name: f.segments[0].file.name,
+            totalParts: f.segments[0].totalParts,
+            groups: f.segments[0].file.groups,
+            firstMessageId: f.segments[0].messageId,
+            poster: f.segments[0].file.poster,
+            segments: f.segments.length,
+            bytes: f.segments.reduce((n, s) => n + s.bytes, 0),
+          })),
+          true
         )
         if (verdict.rule === null) {
-          this.batch.releases.push(chunkOf(key, list, upload))
+          const chunk = chunkOf(key, list, upload)
+          if (!chunk.readable) {
+            // Unnamed: a class the ladder found never names anything stays out of the pending store, its explored
+            // sample excepted (so the rate keeps being measured); counted as garbage (LOG.md (130)).
+            const tool = uploadTool(list[0].segments[0].messageId)
+            const klass = pendingClass(chunk.name, chunk.bytes, chunk.partial === true)
+            if (this.rules.classes.has(classKey(tool, this.options.group, klass)) && !explored(key, this.options.rules.samplePercent)) {
+              this.count('garbage', `${tool} ${this.options.group} #${klass}`, list.reduce((n, f) => n + f.segments.length, 0), list.length)
+              return
+            }
+          }
+          this.batch.releases.push(chunk)
           this.batch.candidates.push(...candidatesOf(key, list, upload))
           return
         }
@@ -259,6 +305,20 @@ export class RangeIndexer {
       else byFile.set(s.fileKey, [s])
     }
     for (const list of byFile.values()) {
+      if (this.options.partialFiles) {
+        // As the Rust window's files(): one article of a multi-part random file is no release; a scrambled name only
+        // before the end, where the rest of the file can still come.
+        const name = list[0].file.name
+        if (list.length === 1 && (list[0].number !== 1 || list[0].totalParts > 1) && (isRandomFile(name) || (!atEnd && isScrambledFile(name)))) {
+          this.count('per-article', uploadTool(list[0].messageId), 1, 0)
+          continue
+        }
+        // One segment per part (two copies of a file share its key here), the first article seen.
+        const parts = new Map<number, EdgeSegment>()
+        for (const s of list.sort((a, b) => a.articleNumber - b.articleNumber)) if (!parts.has(s.number)) parts.set(s.number, s)
+        this.files([{ fileKey: list[0].fileKey, segments: [...parts.values()].sort((a, b) => a.number - b.number) }])
+        continue
+      }
       // One article of a "file" of many parts under a random name: a random subject per article, which
       // never completes from headers (99.8% of the boneless fragments; LOG.md 2026-10-09).
       if (list.length === 1 && isRandomFile(list[0].file.name)) {
@@ -287,7 +347,7 @@ export class RangeIndexer {
         fragment.segments.push([s.number, s.messageId, s.bytes, s.articleNumber])
         if (Number.isFinite(s.postedAt) && (fragment.postedAt === null || s.postedAt < fragment.postedAt)) fragment.postedAt = s.postedAt
       }
-      this.batch.fragments.push(fragment)
+      this.batch.fragments!.push(fragment)
     }
   }
 
@@ -413,8 +473,9 @@ export function chunkOf(key: string, files: AggregatedFile<EdgeSegment>[], uploa
     bytes,
     firstArticle,
     lastArticle,
-    readable: chunkFiles.some((f) => !isRandomFile(f.name)),
+    readable: chunkFiles.some((f) => !isUnreadableName(f.name)),
     hasPar2: chunkFiles.some((f) => /\.par2$/i.test(f.name)),
+    partial: files.some((f) => f.segments.length < f.segments[0].totalParts),
     payload: encodeChunk(chunkFiles),
   }
 }
@@ -436,12 +497,12 @@ export function candidatesOf(key: string, files: AggregatedFile<EdgeSegment>[],
     }
     // An unsplit 7z of more than one segment: its listing is in its last one.
     if (volume === 0 && f.segments.length > 1 && !tail) tail = { volume, file: f }
-    if (isArchiveStart(name)) out.push({ releaseKey: key, upload, kind: 'archive', messageId: f.segments[0].messageId, file: name })
-    else if (/\.par2$/i.test(name) && !isLaterVolume(name)) out.push({ releaseKey: key, upload, kind: 'par2', messageId: f.segments[0].messageId, file: name })
+    if (isArchiveStart(name)) out.push({ releaseKey: key, upload, kind: 'archive', messageId: f.segments[0].messageId, file: name, bytes: f.segments[0].bytes })
+    else if (/\.par2$/i.test(name) && !isLaterVolume(name)) out.push({ releaseKey: key, upload, kind: 'par2', messageId: f.segments[0].messageId, file: name, bytes: f.segments[0].bytes })
   }
-  if (tail) out.push({ releaseKey: key, upload, kind: '7z-tail', messageId: tail.file.segments.at(-1)!.messageId, file: tail.file.segments[0].file.name })
+  if (tail) out.push({ releaseKey: key, upload, kind: '7z-tail', messageId: tail.file.segments.at(-1)!.messageId, file: tail.file.segments[0].file.name, bytes: tail.file.segments.at(-1)!.bytes })
   if (out.length === 0 && files.length === 1 && isRandomFile(files[0].segments[0].file.name)) {
-    out.push({ releaseKey: key, upload, kind: 'lone', messageId: files[0].segments[0].messageId, file: files[0].segments[0].file.name })
+    out.push({ releaseKey: key, upload, kind: 'lone', messageId: files[0].segments[0].messageId, file: files[0].segments[0].file.name, bytes: files[0].segments[0].bytes })
   }
   return out.slice(0, MAX_CANDIDATES)
 }

lib/nntp/index.ts

@@ -1,4 +1,11 @@
-export { XoverClient, MissingArticle, parseOverviews, type Overview, type XoverClientOptions } from './xover_client.js'
+export {
+  XoverClient,
+  MissingArticle,
+  TruncatedReply,
+  parseOverviews,
+  type Overview,
+  type XoverClientOptions,
+} from './xover_client.js'
 export { parseXrefGroups } from './xref.js'
 export { parseGroupResponse, type GroupInfo } from './group.js'
 export { hedge, type Request, type Hedged } from './hedge.js'

lib/nntp/xover_client.ts

@@ -36,6 +36,14 @@ const DOT_LINE = Buffer.from('.\r\n')
  */
 /** The server doesn't have the article (430): taken down or expired. */
 export class MissingArticle extends Error {}
+/**
+ * A compressed reply the server cut short: it ends its compressed output at the first ".\r\n" in it, as if that
+ * were the terminator, and sends nothing more (Eweka; "78 01 2e 0d 0a" for 16 of 14k small XOVERs, LOG.md
+ * 2026-10-25 (105)). The same request without compression answers.
+ */
+export class TruncatedReply extends Error {}
+/** How long a compressed reply that doesn't inflate may stay silent before it counts as cut short. */
+const TRUNCATED_MS = 2_000
 
 export class XoverClient {
   private chunks: Buffer[] = []
@@ -54,6 +62,10 @@ export class XoverClient {
   get buffered() {
     return this.length
   }
+  /** The first `n` bytes received and not consumed yet (what a reply that stalled left; nntp:probe). */
+  bufferedHead(n = 32): Buffer {
+    return this.buffer().subarray(0, n)
+  }
   private lastDataAt = 0
   private longestGap = 0
 
@@ -85,7 +97,11 @@ export class XoverClient {
 
   static async connect(options: XoverClientOptions): Promise<XoverClient> {
     const sock = options.secure
-      ? tls.connect({ host: options.host, port: options.port, servername: options.servername || options.host })
+      ? tls.connect({
+          host: options.host,
+          port: options.port,
+          servername: options.servername || options.host,
+        })
       : net.connect({ host: options.host, port: options.port })
     const client = new XoverClient(sock, options.timeoutMs ?? 120_000)
     // The greeting and login are short: a server holding a connection back (too many open) fails it in 30 s, not timeoutMs.
@@ -94,10 +110,14 @@ export class XoverClient {
       await client.expect(await client.line(), ['200', '201'])
       if (options.user) {
         const reply = await client.command(`AUTHINFO USER ${options.user}`)
-        if (reply.startsWith('381')) await client.expect(await client.command(`AUTHINFO PASS ${options.password ?? ''}`), ['281'])
+        if (reply.startsWith('381'))
+          await client.expect(await client.command(`AUTHINFO PASS ${options.password ?? ''}`), [
+            '281',
+          ])
         else await client.expect(reply, ['281'])
       }
-      if (options.compress) client.compressed = (await client.command('XFEATURE COMPRESS GZIP')).startsWith('290')
+      if (options.compress)
+        client.compressed = (await client.command('XFEATURE COMPRESS GZIP')).startsWith('290')
     } catch (error) {
       client.close()
       throw error
@@ -127,7 +147,9 @@ export class XoverClient {
       const reply = await this.command(`LIST ACTIVE ${pattern}`)
       await this.expect(reply, ['215'])
       // With XFEATURE COMPRESS GZIP on, Eweka compresses this list too ("215 NewsGroups Follow [COMPRESS=GZIP]").
-      const body = reply.includes('COMPRESS=GZIP') ? (await this.compressedRaw(true)).text : await this.plainBody()
+      const body = reply.includes('COMPRESS=GZIP')
+        ? (await this.compressedRaw(true)).text
+        : await this.plainBody()
       return body
         .split('\r\n')
         .filter(Boolean)
@@ -154,7 +176,11 @@ export class XoverClient {
         return deflateAsync(Buffer.from(await this.plainBody()), { level: 1 })
       } finally {
         this.measuring = false
-        this.lastTiming = { replyMs: repliedAt - sentAt, bodyMs: performance.now() - repliedAt, longestGapMs: this.longestGap }
+        this.lastTiming = {
+          replyMs: repliedAt - sentAt,
+          bodyMs: performance.now() - repliedAt,
+          longestGapMs: this.longestGap,
+        }
       }
     })
   }
@@ -180,7 +206,9 @@ export class XoverClient {
     const reply = await this.command(`XOVER ${from}-${to}`)
     if (reply.startsWith('420') || reply.startsWith('423')) return []
     await this.expect(reply, ['224'])
-    const data = reply.includes('COMPRESS=GZIP') ? await this.compressedBody() : await this.plainBody()
+    const data = reply.includes('COMPRESS=GZIP')
+      ? await this.compressedBody()
+      : await this.plainBody()
     return parseOverviews(data)
   }
 
@@ -199,7 +227,8 @@ export class XoverClient {
   }
 
   private async expect(reply: string, codes: string[]) {
-    if (!codes.some((code) => reply.startsWith(code))) throw new Error(`NNTP: unexpected reply "${reply}"`)
+    if (!codes.some((code) => reply.startsWith(code)))
+      throw new Error(`NNTP: unexpected reply "${reply}"`)
   }
 
   private buffer(): Buffer {
@@ -223,10 +252,28 @@ export class XoverClient {
     if (this.failure && this.length === 0) throw this.failure
   }
 
+  /** As more(), but false when nothing came within `ms`. */
+  private async moreWithin(ms: number): Promise<boolean> {
+    if (this.failure) throw this.failure
+    const came = await new Promise<boolean>((resolve) => {
+      const timer = setTimeout(() => resolve(false), ms)
+      this.wake = () => {
+        clearTimeout(timer)
+        resolve(true)
+      }
+    })
+    this.wake = undefined
+    if (this.failure && this.length === 0) throw this.failure
+    return came
+  }
+
   private async line(): Promise<string> {
     while (true) {
       const at = this.buffer().indexOf('\r\n')
-      if (at !== -1) return this.take(at + 2).subarray(0, at).toString('latin1')
+      if (at !== -1)
+        return this.take(at + 2)
+          .subarray(0, at)
+          .toString('latin1')
       await this.more()
     }
   }
@@ -267,7 +314,9 @@ export class XoverClient {
    */
   async bodies(messageIds: string[]): Promise<(Buffer | null)[]> {
     return this.timed(async () => {
-      this.sock.write(messageIds.map((m) => `BODY ${m.startsWith('<') ? m : `<${m}>`}\r\n`).join(''))
+      this.sock.write(
+        messageIds.map((m) => `BODY ${m.startsWith('<') ? m : `<${m}>`}\r\n`).join('')
+      )
       const out: (Buffer | null)[] = []
       for (let i = 0; i < messageIds.length; i++) {
         const reply = await this.line()
@@ -309,7 +358,11 @@ export class XoverClient {
         return ''
       }
       const at = all.indexOf(TERMINATOR, Math.max(0, searched - TERMINATOR.length))
-      if (at !== -1) return this.take(at + TERMINATOR.length).subarray(0, at + 2).toString('utf8').replace(/^\.\./gm, '.')
+      if (at !== -1)
+        return this.take(at + TERMINATOR.length)
+          .subarray(0, at + 2)
+          .toString('utf8')
+          .replace(/^\.\./gm, '.')
       searched = all.length
       await this.more()
     }
@@ -359,7 +412,14 @@ export class XoverClient {
           this.take(all.length)
           return { raw, text: inflated.toString('utf8') }
         } catch {
-          // ".\r\n" inside the compressed data: wait for more.
+          // ".\r\n" inside the compressed data: wait for more, but not for the reply timeout: the server may have
+          // cut its output there (TruncatedReply), and then nothing comes. Raw, the chunk is returned as it is
+          // and reported by the worker; the text is wanted here.
+          if (!(await this.moreWithin(TRUNCATED_MS))) {
+            this.take(all.length)
+            throw new TruncatedReply(`NNTP: compressed reply cut short after ${all.length} bytes`)
+          }
+          continue
         }
       }
       await this.more()

lib/probe/index.ts

@@ -1 +1 @@
-export { inspect, innerName, MAX_NAMES, PROBE_VERSION, type Inspection } from './inspect.js'
+export { inspect, innerName, usableName, MAX_NAMES, PROBE_VERSION, type Inspection } from './inspect.js'

lib/probe/inspect.ts

@@ -5,6 +5,7 @@ import { parseRarHeaders } from '#lib/rar/index'
 import { parseSevenZipTail } from '#lib/sevenzip/index'
 import { parseZipStart } from '#lib/zip/index'
 import { sniffContentType, type ContentType } from '#lib/content/index'
+import { isUnreadableName } from '#lib/release/index'
 
 /** What one probed article showed (REBUILD.md, part 6; stored per release and kind). */
 export interface Inspection {
@@ -25,7 +26,10 @@ export interface Inspection {
    */
   hash16k?: string
   /** A par2 body's recovery set: every file it describes (not kept in release_inspections; par2_files). */
-  par2?: { setId: string; files: { fileId: string; hash16k: string; length: number; name: string }[] }
+  par2?: {
+    setId: string
+    files: { fileId: string; hash16k: string; length: number; name: string }[]
+  }
 }
 
 const HASH_BYTES = 16 * 1024
@@ -63,9 +67,16 @@ function largestFirst(items: { name: string; size: number }[]): string[] {
  */
 export function inspect(body: Buffer, kind?: string): Inspection {
   const segment = decodeBody(body)
-  const result: Inspection = { yencName: segment.meta.begin.name.slice(0, MAX_NAME_LENGTH), size: segment.meta.begin.size, type: sniffContentType(segment.data) }
+  const result: Inspection = {
+    yencName: segment.meta.begin.name.slice(0, MAX_NAME_LENGTH),
+    size: segment.meta.begin.size,
+    type: sniffContentType(segment.data),
+  }
   // The file's first 16 KiB, or all of a smaller file.
-  if (segment.offset === 0 && (segment.data.length >= HASH_BYTES || segment.data.length === result.size)) {
+  if (
+    segment.offset === 0 &&
+    (segment.data.length >= HASH_BYTES || segment.data.length === result.size)
+  ) {
     result.hash16k = createHash('md5').update(segment.data.subarray(0, HASH_BYTES)).digest('hex')
   }
   if (isPar2(segment.data)) {
@@ -76,7 +87,12 @@ export function inspect(body: Buffer, kind?: string): Inspection {
       result.names = largestFirst(set.files.map((f) => ({ name: f.name, size: f.length })))
       result.par2 = {
         setId: set.recoverySetId,
-        files: set.files.map((f) => ({ fileId: f.fileId, hash16k: f.md5_16k, length: f.length, name: f.name.slice(0, MAX_NAME_LENGTH) })),
+        files: set.files.map((f) => ({
+          fileId: f.fileId,
+          hash16k: f.md5_16k,
+          length: f.length,
+          name: f.name.slice(0, MAX_NAME_LENGTH),
+        })),
       }
     }
     return result
@@ -99,7 +115,10 @@ export function inspect(body: Buffer, kind?: string): Inspection {
     }
   }
   // 7z keeps its listing at the end: the last segment of the last volume (a 7z-tail), or a small archive whole.
-  if ((kind === '7z-tail' || result.type === '7z') && segment.offset + segment.data.length === result.size) {
+  if (
+    (kind === '7z-tail' || result.type === '7z') &&
+    segment.offset + segment.data.length === result.size
+  ) {
     const listing = parseSevenZipTail(segment.data)
     if (listing) {
       result.entries = listing.files.length
@@ -110,16 +129,43 @@ export function inspect(body: Buffer, kind?: string): Inspection {
   return result
 }
 
+/**
+ * The usable name one probe of an unnamed release found (the pending ladder, REBUILD.md decision f): the largest
+ * named file of a par2 set or an archive's listing, or a lone file's yEnc name, when it isn't packaging and isn't
+ * random itself (an obfuscated upload's par2 names its random volumes; its yEnc name is mostly the random name
+ * again). Null: this probe named nothing, the next target is tried.
+ */
+export function usableName(inspection: Inspection, kind: string): string | null {
+  const usable = (n: string | undefined) => {
+    if (!n) return null
+    const name = n.replace(/^.*[\\/]/, '').slice(0, 255)
+    return name && !PACKAGING.test(name) && !isUnreadableName(name) ? name : null
+  }
+  const named = (inspection.names ?? []).map(usable).find((n) => n)
+  if (named) return named
+  return kind === 'lone' || !inspection.names?.length ? usable(inspection.yencName) : null
+}
+
 /**
  * A release's best inner name from its inspections (by probe kind): the
  * largest named entry of an archive, else the largest file of a par2 set
  * that isn't packaging, else a lone file's yEnc name. Null when none says
  * more than the release's own name.
  */
-export function innerName(inspections: { kind: string; result: Inspection }[], releaseName = ''): string | null {
+export function innerName(
+  inspections: { kind: string; result: Inspection }[],
+  releaseName = ''
+): string | null {
   const of = (kind: string) => inspections.filter((i) => i.kind === kind).map((i) => i.result)
-  const content = (kind: string) => of(kind).flatMap((i) => i.names ?? []).find((n) => !PACKAGING.test(n))
-  const pick = content('archive') ?? content('7z-tail') ?? content('par2') ?? of('lone').find((i) => i.yencName && !PACKAGING.test(i.yencName))?.yencName
+  const content = (kind: string) =>
+    of(kind)
+      .flatMap((i) => i.names ?? [])
+      .find((n) => !PACKAGING.test(n))
+  const pick =
+    content('archive') ??
+    content('7z-tail') ??
+    content('par2') ??
+    of('lone').find((i) => i.yencName && !PACKAGING.test(i.yencName))?.yencName
   const name = pick?.replace(/^.*[\\/]/, '').slice(0, 255)
   // The release's own name with an extension says nothing new (a random-named file's yEnc name is mostly its random name).
   return name && !name.toLowerCase().startsWith(releaseName.toLowerCase() || '\0') ? name : null

lib/release/classes.ts

@@ -0,0 +1,55 @@
+/**
+ * The class of an unnamed release (REBUILD.md decision f; LOG.md (130)): what the headers say about its chance of a
+ * name, from the pending ladder's own record. Size bucket × name shape × partial, 1..70, kept in step with
+ * names.rs pending_class and the SQL pending_class() (migration 6800): the batch store sets it with the SQL, the
+ * edge skips the classes a lease marks dropped.
+ */
+
+/** Size buckets: under 1 MB, 10 MB, 100 MB, 1 GB, and from 1 GB. */
+export const SIZE_BUCKETS = [1_000_000, 10_000_000, 100_000_000, 1_000_000_000]
+/** Name shapes, 1..7: hex, digits, base32, letters only, mixed letters and digits, with separators, other. */
+export const SHAPES = [
+  'hex',
+  'digits',
+  'base32',
+  'letters',
+  'alnum',
+  'separators',
+  'other',
+] as const
+
+export function sizeBucket(bytes: number): number {
+  let i = 0
+  while (i < SIZE_BUCKETS.length && bytes >= SIZE_BUCKETS[i]) i++
+  return i
+}
+
+export function nameShape(name: string): number {
+  const s = name.replace(/\.[A-Za-z0-9]{1,4}$/, '')
+  if (/^[0-9a-fA-F]{16,}$/.test(s)) return 1
+  if (/^[0-9]{6,}$/.test(s)) return 2
+  if (/^[a-z2-7]{20,}$/.test(s) && /[2-7]/.test(s)) return 3
+  if (/^[A-Za-z]+$/.test(s)) return 4
+  if (/^[A-Za-z0-9]+$/.test(s)) return 5
+  if (/^[A-Za-z0-9._-]+$/.test(s)) return 6
+  return 7
+}
+
+/** 1..70: (size bucket × 7 + shape − 1) × 2 + partial + 1. */
+export function pendingClass(name: string, bytes: number, partial: boolean): number {
+  return sizeBucket(bytes) * 14 + (nameShape(name) - 1) * 2 + (partial ? 1 : 0) + 1
+}
+
+/** A class as droppedClasses lists it: "tool\tgroup\tklass". */
+export function classKey(tool: string, group: string, klass: number): string {
+  return `${tool}\t${group}\t${klass}`
+}
+
+/** A class in words ("10-100M hex partial"). */
+export function describeClass(klass: number): string {
+  const k = klass - 1
+  const size = Math.floor(k / 14)
+  const shape = SHAPES[Math.floor((k % 14) / 2)]
+  const sizes = ['<1M', '1-10M', '10-100M', '100M-1G', '1G+']
+  return `${sizes[size]} ${shape}${k % 2 ? ' partial' : ''}`
+}

lib/release/identity.ts

@@ -105,6 +105,29 @@ export function subjectRemainder(subject: string): string {
     .trim()
 }
 
+/**
+ * A subject name as shown: brackets the counters left unmatched removed ("… Facialized 002/477] - 0177.jpg", the
+ * counter's "[" missing, gave "… Facialized ]"; LOG.md 2026-10-29 (121)). Only the name: the key keeps the remainder.
+ */
+export function displayName(remainder: string): string {
+  if (!/[[\]()]/.test(remainder)) return remainder
+  const drop = new Set<number>()
+  for (const [open, close] of [['[', ']'], ['(', ')']]) {
+    const opens: number[] = []
+    for (let i = 0; i < remainder.length; i++) {
+      if (remainder[i] === open) opens.push(i)
+      else if (remainder[i] === close) {
+        if (opens.length > 0) opens.pop()
+        else drop.add(i)
+      }
+    }
+    for (const i of opens) drop.add(i)
+  }
+  if (drop.size === 0) return remainder
+  const name = remainder.split('').filter((_, i) => !drop.has(i)).join('').replace(/ {2,}/g, ' ').trim()
+  return name || remainder
+}
+
 /**
  * Posting tools put random per-file tokens and poster IDs in subjects (e.g.
  * `elys-de81bcabef5faf6a - [Nyh5Q07Y]`); only text that still reads like a
@@ -448,7 +471,7 @@ export function releaseIdentity(input: ReleaseInput): ReleaseIdentity | null {
     if (!image && RANDOM_VOLUME.test(input.name) && /^[A-Za-z0-9_-]+$/.test(stem) && looksRandom(stem.replace(/[-_]/g, ''))) parts.push(stem)
     return {
       key: hash(...parts),
-      name: remainder,
+      name: displayName(remainder),
       fileTotal,
       source: 'subject',
     }
@@ -456,7 +479,7 @@ export function releaseIdentity(input: ReleaseInput): ReleaseIdentity | null {
   // Not a name, but a post id (e.g. `83110629-n [10/18] - "random"`): with the
   // [i/M] counter it groups the files of one post whose names are random.
   if (!tag && postTotal && isPostToken(remainder)) {
-    return { key: hash('post', poster, remainder, postTotal), name: remainder, fileTotal, source: 'post' }
+    return { key: hash('post', poster, remainder, postTotal), name: displayName(remainder), fileTotal, source: 'post' }
   }
   // A scene music post's index files under a bare counter go with its tracks: the artist and the total ("[01/13] -
   // "00-ammar-turk_uit_de_kroeg-single-web-nl-2022-cover.jpg"" with "[05/13] - "01-ammar-turk_uit_de_kroeg.mp3"": the

lib/release/index.ts

@@ -1,5 +1,7 @@
 /**
  * Grouping of files into releases (downloads), independent of AdonisJS.
  */
-export { releaseIdentity, par2ReleaseIdentity, par2ReleaseName, rarReleaseIdentity, isGenericName, fileStem, sizeAfterName, runOn, subjectRemainder, namesRelease, isPostToken, type ReleaseIdentity, type ReleaseInput } from './identity.js'
+export { releaseIdentity, displayName, par2ReleaseIdentity, par2ReleaseName, rarReleaseIdentity, isGenericName, fileStem, sizeAfterName, runOn, subjectRemainder, namesRelease, isPostToken, type ReleaseIdentity, type ReleaseInput } from './identity.js'
 export { looksRandom } from './random.js'
+export { isUnreadableName, tokenKind } from './readable.js'
+export { pendingClass, classKey, describeClass, nameShape, sizeBucket } from './classes.js'

lib/release/keys.ts

@@ -0,0 +1,34 @@
+import { createHash } from 'node:crypto'
+
+/**
+ * The catalog's keys as voyager 3 stores them (migration 6000): 20-byte SHA-1s in bytea columns.
+ * - identity_key: the identity's hash (lib/release/identity.ts gives it as 40 hex characters).
+ * - release_key: one upload of an identity, unique: the identity, the owner group and the first article the window
+ *   saw of it. The same release sent twice (a tail release provisionally, then finished) hits the same row; a repost
+ *   starts at another article and is a release of its own (decision d). Deterministic, so no numbering under a lock.
+ */
+const HEX40 = /^[0-9a-f]{40}$/i
+
+function sha1(text: string): Buffer {
+  return createHash('sha1').update(text).digest()
+}
+
+/** The identity key's bytes: 40 hex characters decoded, any other text (a spot's, a test's) hashed. */
+export function identityKey(key: string): Buffer {
+  return HEX40.test(key) ? Buffer.from(key, 'hex') : sha1(key)
+}
+
+/** The release key's bytes for an identity in a group, first seen at an article (null for releases without one: a spot's). */
+export function releaseKey(
+  identity: string,
+  groupId: number | null,
+  firstArticle: number | null
+): Buffer {
+  return sha1(`release\n${identity}\n${groupId ?? ''}\n${firstArticle ?? ''}`)
+}
+
+/** A stored key as hex, for logs, the API and release_merges.from_key. */
+export function keyHex(key: Buffer | Uint8Array | string | null | undefined): string | null {
+  if (key === null || key === undefined) return null
+  return typeof key === 'string' ? key : Buffer.from(key).toString('hex')
+}

lib/release/readable.ts

@@ -0,0 +1,148 @@
+import { fileStem } from './identity.js'
+import { looksRandom } from './random.js'
+
+/**
+ * Readable or not: whether a name says what a release is (the catalog's readable flag, the prober's inner names).
+ * looksRandom (random.ts) knows the single-token shapes; this reads a name token by token, so separators, spaces
+ * and random "extensions" don't hide a scramble (LOG.md (129)).
+ */
+/**
+ * Tokens that don't make a name readable on their own: packaging, volume numbers and the posting tools' fixed
+ * words ("Description - Lt5nHjlo3RiuBb3C", "bd5b4a59f3aa87114dcf75c0-addon").
+ */
+const NOISE =
+  /^(addon|samples?|proof|subs?|nfo|part\d+|vol\d+|description|repost|rep|extras?|bonus|cd\d*|dvd\d*|dis[ck]\d*|par2|rar|zip|7z|mkv|mp4|avi|vol)$/i
+/** Mixed tokens that are codes, not scrambles: S02E10, 1080p, 10bit, x264, Win10x64, BDRip720p, 3DMark03, v2, 5th. */
+const CODE =
+  /^([sS]\d{1,2}[eE]\d{1,3}|[xh]26[45]|[A-Za-z]+\d+([a-z]{1,3}|x(64|86))?|\d+[A-Za-z]{1,3}|\d+[A-Za-z]+\d+)$/
+/** The NZBSPLIT tools' hash between their markers is no part of the name. */
+const NZBSPLIT = /__NZBSPLIT__[0-9a-f]{32}__NZBSPLIT__/i
+
+/** Scene tags write "i" for "I" between capitals (XCiTE, PRESCRiPTiON): read it as a capital there. */
+function sceneCase(t: string): string {
+  let out = ''
+  for (let i = 0; i < t.length; i++) {
+    out +=
+      t[i] === 'i' && i > 0 && i < t.length - 1 && /[A-Z]/.test(t[i - 1]) && /[A-Z]/.test(t[i + 1])
+        ? 'I'
+        : t[i]
+  }
+  return out
+}
+
+/**
+ * The signals of a scrambled token: letter/digit and lower-to-upper switches; case switches in both directions; lone
+ * lowercase letters between capitals ("NtpUgCmS"); an inner run of 2+ capitals with lowercase on both sides
+ * ("tBWTuz", "YDoaROv"); the longest consonant run; vowels (a lowercase y too: "McCartney"); runs of one character class.
+ */
+function signals(t: string) {
+  let switches = 0
+  let caseSwitches = 0
+  let lone = 0
+  let consonants = 0
+  let maxConsonants = 0
+  for (let i = 0; i < t.length; i++) {
+    const c = t[i]
+    if (/[b-df-hj-np-tv-xz]/i.test(c)) maxConsonants = Math.max(maxConsonants, ++consonants)
+    else consonants = 0
+    if (i === 0) continue
+    const a = t[i - 1]
+    if (/\d/.test(a) !== /\d/.test(c)) switches++
+    else if (/[a-z]/.test(a) && /[A-Z]/.test(c)) switches++
+    if ((/[a-z]/.test(a) && /[A-Z]/.test(c)) || (/[A-Z]/.test(a) && /[a-z]/.test(c))) caseSwitches++
+    if (
+      i < t.length - 1 &&
+      /[a-z]/.test(c) &&
+      c !== 'i' &&
+      /[A-Z]/.test(a) &&
+      /[A-Z]/.test(t[i + 1])
+    )
+      lone++
+  }
+  const innerUpper = /[a-z][A-Z]{2,}[a-z]/.test(t)
+  const vowels = (t.match(/[aeiouAEIOU]|y/g) ?? []).length
+  const runs = (t.match(/[a-z]+|[A-Z]+|\d+/g) ?? []).length
+  const letters = (t.match(/[A-Za-z]/g) ?? []).length
+  const maxLower = Math.max(0, ...(t.match(/[a-z]+/g) ?? []).map((r) => r.length))
+  return {
+    switches,
+    caseSwitches,
+    lone,
+    innerUpper,
+    maxConsonants,
+    vowels,
+    runs,
+    letters,
+    maxLower,
+  }
+}
+
+/**
+ * What one alphanumeric token of a name is: a word (letters only, in one case or Capitalized, 3+ long, not a packaging
+ * word), a scramble, or neither (digits, codes, CamelCase, short tokens). Kept in step with names.rs token_kind.
+ */
+export function tokenKind(t: string): 'word' | 'random' | null {
+  if (/^\d+$/.test(t) || NOISE.test(t) || CODE.test(t)) return null
+  if (/^[A-Za-z]+$/.test(t)) {
+    const s = signals(sceneCase(t))
+    if (s.caseSwitches === 0 || (s.caseSwitches === 1 && /^[A-Z][a-z]+$/.test(t))) {
+      // A base32 consonant soup ("mrrtmzrzmi4dgmrzmyytqmlbztczbx" without its digit) is no word.
+      if (t.length >= 10 && s.vowels * 6 < t.length) return 'random'
+      return t.length >= 3 ? 'word' : null
+    }
+    // Mixed case beyond Capitalization: CamelCase (neutral) or a scramble. Two CamelCase words with few vowels
+    // (FastFilms, BestCrypt, SchiffsModell) have four runs and no lone letter; a scramble has more runs, a lone
+    // letter or an inner capital run.
+    if (
+      t.length >= 6 &&
+      s.caseSwitches >= 2 &&
+      s.vowels * 4 < t.length &&
+      (s.runs >= 5 || s.lone >= 1 || s.innerUpper)
+    )
+      return 'random'
+    return null
+  }
+  if (t.length >= 8 && /^[0-9a-f]+$/i.test(t) && (t.match(/\d/g)?.length ?? 0) >= 2) return 'random'
+  if (/^[A-Z0-9]+$/.test(t)) {
+    // 51294P1CUKR2R35X; COD4MW2REMASTERED (4 switches) and BEATLES1964REMASTER (2) stay.
+    return t.length >= 12 && signals(t).switches >= 6 ? 'random' : null
+  }
+  // Words with numbers between them (Laura15Juli2026, Back4WinXP): every letter run a word or an acronym.
+  if (
+    (t.match(/[A-Za-z]+/g) ?? []).every(
+      (w) =>
+        /^[A-Z]{2,}$/.test(w) ||
+        (/^[A-Z]?[a-z]{2,}$/.test(w) && (w.match(/[aeiouy]/g) ?? []).length * 5 >= w.length)
+    )
+  )
+    return null
+  const s = signals(sceneCase(t))
+  if (t.length >= 5 && s.switches >= 3) return 'random'
+  if (t.length >= 6 && s.switches >= 2 && (s.lone >= 1 || s.innerUpper || s.maxConsonants >= 4))
+    return 'random'
+  // Few vowels and no lowercase run of 3+ ("Jk5Pud", "RKi9BCg"; not "HDR10Plus").
+  if (t.length >= 6 && s.switches >= 2 && s.vowels * 4 < s.letters && s.maxLower <= 2)
+    return 'random'
+  return null
+}
+
+/**
+ * A file or release name no one can read: looksRandom's shapes, or more scrambled tokens than words by letter count
+ * ("Zx5AjNaK-mh7CFg6Hf8uu", "Description - Lt5nHjlo3RiuBb3C", "PsNdnYiugfZ"; not "Paul McCartney - Band on the Run",
+ * "ShesSoWild - Ester Ana 1080p"). Separators and a random "extension" of any length don't hide a scramble, which
+ * looksRandom's single-token shapes miss (15% of the readable catalog on 2026-11-02; LOG.md (129)). The readable
+ * flag; the drop rules keep looksRandom, which is narrower.
+ */
+export function isUnreadableName(name: string): boolean {
+  const stem = fileStem(name).replace(NZBSPLIT, ' ')
+  if (looksRandom(stem)) return true
+  let words = 0
+  let random = 0
+  for (const t of stem.split(/[^A-Za-z0-9]+/)) {
+    if (!t) continue
+    const kind = tokenKind(t)
+    if (kind === 'word') words += t.length
+    else if (kind === 'random') random += t.length
+  }
+  return random > 0 && random > words
+}

lib/release/source.ts

@@ -0,0 +1,38 @@
+/**
+ * How a release got its name, as releases.source stores it (a smallint; voyager 3, migration 6000). The names are
+ * what the edge, the spot inspector and the API use; the numbers never change meaning once written.
+ */
+export const SOURCES = {
+  /** A name the subject announces ("name" yEnc (k/N)). */
+  subject: 1,
+  /** The post's files, named after the first file (no release name in the subject). */
+  post: 2,
+  /** The files' shared stem. */
+  stem: 3,
+  /** A Spotnet spot's files that matched no release. */
+  spot: 4,
+  /** An NZB's files. */
+  nzb: 5,
+  /** One poster's files that differ only by a number sequence, or share a prefix (voyager 3). */
+  album: 6,
+  /** An obfuscated post's stem, read from the yEnc body (identity.ts: weak). */
+  'weak-stem': 7,
+} as const
+
+export type Source = keyof typeof SOURCES
+
+const NAMES = Object.fromEntries(
+  Object.entries(SOURCES).map(([name, code]) => [code, name])
+) as Record<number, Source>
+
+/** The code of a source name; throws for an unknown one (a typo would write a wrong number for good). */
+export function sourceCode(name: string): number {
+  const code = SOURCES[name as Source]
+  if (code === undefined) throw new Error(`Unknown release source "${name}"`)
+  return code
+}
+
+/** The name of a stored code; "unknown" for one not listed (never written, but a row might be read by an older build). */
+export function sourceName(code: number | string | null | undefined): Source | 'unknown' {
+  return NAMES[Number(code)] ?? 'unknown'
+}

lib/release/text_kind.ts

@@ -0,0 +1,60 @@
+/**
+ * What a searchable string of a release is, as release_texts.kind stores it (a smallint; voyager 3, migration 6200).
+ * The table holds every string search matches against (the release name too), appended by whoever learns one: the
+ * batch store the name, the prober the file names inside an archive, later passes NFO titles, par2 set names,
+ * media titles. The numbers never change meaning once written.
+ */
+export const TEXT_KINDS = {
+  /** The release name (releases.name, once more for the one trigram index). */
+  name: 1,
+  /** A file name inside an archive (the prober's inner names). */
+  file: 2,
+  /** A title read from an NFO. */
+  nfo: 3,
+  /** A par2 set's name when it differs from the release name. */
+  par2: 4,
+  /** A title from a media container's metadata. */
+  media: 5,
+  /** An album's member, by its own file name (the edge's albums: one poster's files differing by a number sequence). */
+  member: 6,
+} as const
+
+export type TextKind = keyof typeof TEXT_KINDS
+
+/** The longest string stored (release_texts.text is a varchar(255)); longer ones are cut, not refused. */
+export const TEXT_MAX = 255
+/** The most strings one writer appends per release and kind (an archive of two thousand files is not two thousand rows). */
+export const TEXTS_PER_KIND = 20
+
+const NAMES = Object.fromEntries(
+  Object.entries(TEXT_KINDS).map(([name, code]) => [code, name])
+) as Record<number, TextKind>
+
+/** The code of a kind; throws for an unknown one (a typo would write a wrong number for good). */
+export function textKindCode(name: string): number {
+  const code = TEXT_KINDS[name as TextKind]
+  if (code === undefined) throw new Error(`Unknown release text kind "${name}"`)
+  return code
+}
+
+/** The name of a stored code; "unknown" for one not listed. */
+export function textKindName(code: number | string | null | undefined): TextKind | 'unknown' {
+  return NAMES[Number(code)] ?? 'unknown'
+}
+
+/**
+ * The strings a writer appends for one kind: trimmed, cut to TEXT_MAX, empty ones and repeats dropped, the first
+ * TEXTS_PER_KIND kept in the order given (the writer puts the most telling first).
+ */
+export function textsToStore(texts: Iterable<string | null | undefined>): string[] {
+  const seen = new Set<string>()
+  const out: string[] = []
+  for (const raw of texts) {
+    const text = (raw ?? '').trim().slice(0, TEXT_MAX)
+    if (!text || seen.has(text)) continue
+    seen.add(text)
+    out.push(text)
+    if (out.length === TEXTS_PER_KIND) break
+  }
+  return out
+}

lib/rules/index.ts

@@ -1,2 +1,2 @@
 export { uploadTool, signatureGroup, signatureKey, explored } from './upload_tools.js'
-export { RuleSet, DEFAULT_RULES, isRandomFile, ownerOf, posterWords, type Rules, type Verdict } from './rule_set.js'
+export { RuleSet, DEFAULT_RULES, isRandomFile, isScrambledFile, ownerOf, posterWords, stubArchiveLegs, isNonsensePiece, isTinyArchive, STUB_ARTICLE_BYTES, STUB_MIN_BYTES, STUB_MAX_BYTES, STUB_WIDE_GROUPS, NONSENSE_MIN_BYTES, NONSENSE_MAX_BYTES, TINY_MIN_BYTES, type Rules, type Verdict, type JudgedFile } from './rule_set.js'

lib/rules/rule_set.ts

@@ -12,6 +12,11 @@ export interface Rules {
   priorities: Record<string, number>
   /** Dropped signatures, as "tool\tgroup" (GarbageFilter's table). */
   droppedSignatures: string[]
+  /**
+   * Dropped classes of unnamed releases, as "tool\tgroup\tklass" (lib/release/classes.ts; the pending ladder's record,
+   * LOG.md (130)): the edge keeps them out of the pending store, the explored sample excepted.
+   */
+  droppedClasses: string[]
   /** Tools whose random-named uploads are dropped in every group (named < 1%; LOG.md 2026-10-08 (5)). */
   droppedTools: string[]
   /** A random-named file of at most this many parts: a random subject per article, needs bodies. */
@@ -30,6 +35,24 @@ export interface Rules {
    * again under other titles (119k releases on 2026-10-16; LOG.md 2026-10-16 (65)).
    */
   droppedPosters: string[]
+  /**
+   * Drop settled releases of one unsplit archive without a par2, posted to 2+ groups, under STUB_MAX_BYTES, whose
+   * articles are at most STUB_ARTICLE_BYTES with the release at least STUB_MIN_BYTES, or posted to 4+ groups, or by
+   * a webmail poster: the exe-stub and recycled-installer floods under film and software names (55% of the catalog
+   * on 2026-10-30; 99.3% of the known fakes, 0 known content; LOG.md (123)-(126), (128)).
+   */
+  dropStubArchives: boolean
+  /**
+   * Drop one-article archives of NONSENSE_MIN_BYTES to NONSENSE_MAX_BYTES under a release-like, unscrambled name:
+   * the newznzb.* pieces (9% of the catalog; LOG.md (123), (125)).
+   */
+  dropNonsensePieces: boolean
+  /**
+   * Drop one-article archives of TINY_MIN_BYTES to NONSENSE_MIN_BYTES under a release-like name (spaces allowed),
+   * outside the exempt groups, without an e-book, subtitle, crack or sidecar word: the 2015-17 exe-stub flood in
+   * miniature (a RAR holding "<title>.exe" at 2-60 kB; 1.65M catalog rows; LOG.md (131), (133)).
+   */
+  dropTinyArchives: boolean
 }
 
 export const DEFAULT_RULES: Omit<Rules, 'priorities' | 'droppedSignatures'> = {
@@ -39,9 +62,37 @@ export const DEFAULT_RULES: Omit<Rules, 'priorities' | 'droppedSignatures'> = {
   samplePercent: 1,
   dropExecutables: true,
   droppedPosters: ['celebmovie', 'celebwow', 'celebfx', 'iluvcelebs', 'nudecelebs', 'celebsextapes'],
+  dropStubArchives: true,
+  dropNonsensePieces: true,
+  dropTinyArchives: true,
+  droppedClasses: [],
 }
 
-export type Verdict = { rule: null } | { rule: 'poster' | 'executable' | 'garbage' | 'per-article'; key: string; sample: boolean }
+/** The stub-archive rule's bounds (edge/src/names.rs): Power-Post's 250 kB default article, the fakes' size span. */
+export const STUB_ARTICLE_BYTES = 262144
+export const STUB_MIN_BYTES = 1_000_000
+export const STUB_MAX_BYTES = 100_000_000
+export const STUB_WIDE_GROUPS = 4
+/** The nonsense-piece rule's bounds: under 100 kB are cue/log sidecars and adverts, 1 MB and up is rule 1's. */
+export const NONSENSE_MIN_BYTES = 100_000
+export const NONSENSE_MAX_BYTES = 1_000_000
+/** The tiny-archive rule's floor: a RAR holding one exe stub is 2 kB and up; below are empty archives and nfo rars. */
+export const TINY_MIN_BYTES = 2_000
+
+export type Verdict =
+  | { rule: null }
+  | { rule: 'poster' | 'executable' | 'garbage' | 'per-article' | 'stub-archive' | 'nonsense-piece' | 'tiny-archive'; key: string; sample: boolean }
+
+/** A file as the judge sees it: its name, parts, groups, poster, and the articles seen with their bytes. */
+export interface JudgedFile {
+  name: string
+  totalParts: number
+  groups: string[]
+  firstMessageId: string
+  poster: string
+  segments: number
+  bytes: number
+}
 
 /** A poster as droppedPosters matches it: lowercased, letters and digits only. */
 export function posterWords(poster: string): string {
@@ -51,11 +102,117 @@ export function posterWords(poster: string): string {
 /** Files that come along with a release's content: they don't make it one of executables or not. */
 const SIDECAR = /\.(par2|nfo|sfv|txt|jpe?g|png|nzb)$/i
 const EXECUTABLE = /\.(exe|scr|com|pif|bat|cmd|vbs|lnk)\s*$/i
+/** One zip, rar or 7z that is not a volume of a set. */
+const ARCHIVE = /\.(zip|rar|7z)$/i
+/** "cneg" is "part" in ROT13: the a.b.erotica posts named that way are volumes of passworded videos (LOG.md (128)). */
+const VOLUME = /\.((part|cneg)\d+\.rar|r\d\d|z\d\d|7z\.\d\d\d|\d\d\d)$/i
+/** A poster at a free mail provider: the stub floods' throwaway accounts. */
+const WEBMAIL = /@(gmail|googlemail|yahoo|hotmail|outlook|live|msn|aol|gmx|web|t-online|freenet|mail|yandex|icloud|me|hushmail|protonmail|libero|orange|free|arcor|wanadoo|ymail|rocketmail|inbox|mail\.ru|bk|list)\./i
+/** A name that reads as a release: a year, a scene tag, a group suffix or dotted words. */
+const RELEASE_LIKE = /(\b(19|20)\d\d\b|[._-](720p|1080p|2160p|x264|x265|hevc|bluray|web-?dl|webrip|hdtv|xvid|s\d\de\d\d|german|multi|crack|repack|proper|flac|mp3|xxx|mp4|retail|keygen|portable|incl|v\d+(\.\d+)+)\b|-[A-Za-z0-9]{2,}$|\.[A-Za-z0-9]+\.[A-Za-z0-9]+\.)/i
+
+/**
+ * Groups whose communities post real single small archives (karaoke zips, e-books, comics, tabs, fonts) the way the
+ * stub floods do (Power-Post defaults, a crosspost or two): a release touching one is exempt from the stub-archive
+ * rule. Costs 16 of 32,308 known fakes in the 217k sample (LOG.md (127)).
+ */
+const STUB_EXEMPT_GROUPS = /(sound|mp3|karaoke|music|audio|flac|lossless|e-?book|epub|comic|guitar|font|magazine)/i
+
+/** The single unsplit archive a release consists of, or null. */
+function singleArchive(files: JudgedFile[]): JudgedFile | null {
+  return files.length === 1 && ARCHIVE.test(files[0].name) && !VOLUME.test(files[0].name) ? files[0] : null
+}
+
+/**
+ * The legs of the stub-archive rule a settled single archive under STUB_MAX_BYTES (the fakes stop at ~90 MB;
+ * hash-named uploads, real albums and passworded videos above it don't) trips: "A" (articles <= STUB_ARTICLE_BYTES,
+ * release >= STUB_MIN_BYTES: a one-article release always has small articles, so e-books and sidecars are exempt),
+ * "X" (4+ groups) and "W" (webmail poster). Empty when none.
+ */
+export function stubArchiveLegs(f: JudgedFile): string {
+  if (f.groups.length < 2 || f.bytes >= STUB_MAX_BYTES || f.groups.some((g) => STUB_EXEMPT_GROUPS.test(g))) return ''
+  let legs = ''
+  if (f.bytes / Math.max(f.segments, 1) <= STUB_ARTICLE_BYTES && f.bytes >= STUB_MIN_BYTES) legs += 'A'
+  if (f.groups.length >= STUB_WIDE_GROUPS) legs += 'X'
+  if (WEBMAIL.test(f.poster)) legs += 'W'
+  return legs
+}
+
+/** An e-book by its name ("Author - Title (2020) (epub).rar"): one small article is a book, not a nonsense piece. */
+const EBOOK = /\b(epub|mobi|azw3?|pdf|djvu|cbr|cbz|e-?books?)\b/i
+
+/**
+ * A nonsense piece: one article, one small archive under a release-like, unscrambled name without spaces (the
+ * newznzb.* pieces are scene-styled or random with a "-XX" suffix; books and albums are written with spaces).
+ */
+export function isNonsensePiece(f: JudgedFile): boolean {
+  return (
+    f.totalParts === 1 &&
+    f.segments === 1 &&
+    f.bytes >= NONSENSE_MIN_BYTES &&
+    f.bytes < NONSENSE_MAX_BYTES &&
+    !f.name.includes(' ') &&
+    RELEASE_LIKE.test(fileStem(f.name)) &&
+    !EBOOK.test(f.name) &&
+    !isScrambledFile(f.name) &&
+    !isRandomFile(f.name)
+  )
+}
+
+/** RELEASE_LIKE with spaces as separators too: the tiny fakes are posted under names written either way. */
+const RELEASE_LIKE_SPACED = new RegExp(RELEASE_LIKE.source.replace('[._-]', '[._ -]'), 'i')
+
+/**
+ * Words of the real tiny archives under release-like names: subtitles, cracks, trainers and keygens (a few kB of
+ * exe is the content), covers and sidecars.
+ */
+const TINY_CONTENT =
+  /\b(subs?|subtitles?|subpack|srt|untertitel|ondertitels?|ondertiteling|legendas?|napisy|sottotitoli|cracks?|cracked|keygen|keymaker|trainer|patch|patched|serials?|activator|loader|nocd|no-cd|fix|covers?|artwork|booklet|scans?|proof|sample|info|readme)\b/i
+
+/**
+ * A tiny archive: one article of TINY_MIN_BYTES to NONSENSE_MIN_BYTES holding one archive under a release-like,
+ * unscrambled name (spaces allowed), outside the groups that post real small archives, and without an e-book,
+ * subtitle, crack or sidecar word: the 2015-17 exe-stub flood in miniature (LOG.md (131)).
+ */
+export function isTinyArchive(f: JudgedFile): boolean {
+  return (
+    f.totalParts === 1 &&
+    f.segments === 1 &&
+    f.bytes >= TINY_MIN_BYTES &&
+    f.bytes < NONSENSE_MIN_BYTES &&
+    !f.groups.some((g) => STUB_EXEMPT_GROUPS.test(g)) &&
+    RELEASE_LIKE_SPACED.test(fileStem(f.name)) &&
+    !EBOOK.test(f.name) &&
+    !TINY_CONTENT.test(f.name) &&
+    !isScrambledFile(f.name) &&
+    !isRandomFile(f.name)
+  )
+}
 
 export function isRandomFile(name: string) {
   return looksRandom(fileStem(name))
 }
 
+/**
+ * At least half the letters and digits of a file's stem in tokens of 5 or more that switch case (lower to upper) or
+ * between letters and digits 3 times or more ("nYDeXeF49dieZ1J7flK-JFzZ3j.ZEDsr", "bslXlwgPiqZadhRa"): the names a
+ * random-subject-per-article upload gives each article that looksRandom misses for their separators.
+ */
+export function isScrambledFile(name: string) {
+  let total = 0
+  let scrambled = 0
+  for (const token of fileStem(name).split(/[^A-Za-z0-9]+/)) {
+    total += token.length
+    let switches = 0
+    for (let i = 1; i < token.length; i++) {
+      const [a, b] = [token[i - 1], token[i]]
+      if ((/[a-z]/.test(a) && /[A-Z]/.test(b)) || /\d/.test(a) !== /\d/.test(b)) switches++
+    }
+    if (token.length >= 5 && switches >= 3) scrambled += token.length
+  }
+  return total >= 8 && scrambled * 2 >= total
+}
+
 /** The owner of an article posted to `groups` (its Xref), or null when none of them is indexed. */
 export function ownerOf(groups: string[], priorities: Record<string, number>): string | null {
   let owner: string | null = null
@@ -68,20 +225,24 @@ export function ownerOf(groups: string[], priorities: Record<string, number>): s
 
 export class RuleSet {
   private signatures: Set<string>
+  readonly classes: Set<string>
   private tools: Set<string>
   private posters: string[]
   constructor(readonly rules: Rules) {
     this.posters = rules.droppedPosters ?? []
     this.signatures = new Set(rules.droppedSignatures)
+    this.classes = new Set(rules.droppedClasses ?? [])
     this.tools = new Set(rules.droppedTools)
   }
 
   /**
    * A release's verdict from its files: dropped when its poster is; when its content is bare
-   * executables; else kept as a whole when any file has a readable name; else dropped as garbage when its signature or tool is
-   * dropped, as per-article random subjects when every file has few parts.
+   * executables; when it is a settled single archive that is a stub or a nonsense piece; else kept as a whole when
+   * any file has a readable name; else dropped as garbage when its signature or tool is dropped, as per-article
+   * random subjects when every file has few parts. `settled`: the window closed on the release (a par2 can still
+   * join an open one), which the single-archive rules need.
    */
-  judge(releaseKey: string, files: { name: string; totalParts: number; groups: string[]; firstMessageId: string; poster: string }[]): Verdict {
+  judge(releaseKey: string, files: JudgedFile[], settled = true): Verdict {
     const sample = explored(releaseKey, this.rules.samplePercent)
     if (this.posters.length > 0 && files.length > 0) {
       const words = posterWords(files[0].poster)
@@ -94,6 +255,15 @@ export class RuleSet {
         return { rule: 'executable', key: EXECUTABLE.exec(content[0].name)![1].toLowerCase(), sample }
       }
     }
+    const archive = settled ? singleArchive(files) : null
+    if (archive) {
+      if (this.rules.dropStubArchives) {
+        const legs = stubArchiveLegs(archive)
+        if (legs !== '') return { rule: 'stub-archive', key: legs, sample }
+      }
+      if (this.rules.dropNonsensePieces && isNonsensePiece(archive)) return { rule: 'nonsense-piece', key: uploadTool(archive.firstMessageId), sample }
+      if (this.rules.dropTinyArchives && isTinyArchive(archive)) return { rule: 'tiny-archive', key: uploadTool(archive.firstMessageId), sample }
+    }
     if (files.some((f) => !isRandomFile(f.name))) return { rule: null }
     const first = files[0]
     const tool = uploadTool(first.firstMessageId)

public/index.html

@@ -22,7 +22,7 @@
           <label class="search-box">
             <svg aria-hidden="true" viewBox="0 0 24 24"><circle cx="11" cy="11" r="7" /><path d="m20 20-3.5-3.5" /></svg>
             <input id="q" type="search" autocomplete="off" spellcheck="false"
-                   placeholder="Search release names…" aria-label="Search" />
+                   placeholder="Search release names… (^word: at a word's start)" aria-label="Search" />
             <kbd aria-hidden="true">/</kbd>
           </label>
           <label class="select">

start/env.ts

@@ -74,6 +74,14 @@ export default await Env.create(new URL('../', import.meta.url), {
   SPOT_STANDALONE_AFTER_HOURS: Env.schema.number.optional(),
   CORPUS_AHEAD: Env.schema.number.optional(),
   PROBE_PIPELINE: Env.schema.number.optional(),
+  SNAPSHOT_DIR: Env.schema.string.optional(),
+  LEASE_MIN_PRIORITY: Env.schema.number.optional(),
+  LEASE_MAX_PRIORITY: Env.schema.number.optional(),
+  LEASE_CAPTURE_ONLY: Env.schema.boolean.optional(),
+  STREAM_CHUNK_WAIT_MINUTES: Env.schema.number.optional(),
+  PENDING_SHARE: Env.schema.number.optional(),
+  PENDING_MIN_PROBES: Env.schema.number.optional(),
+  PENDING_MAX_NAMED: Env.schema.number.optional(),
   SPOOL_AHEAD_PER_LEASE: Env.schema.number.optional(),
   SPOOL_MIN_AHEAD: Env.schema.number.optional(),
   SPOOL_MAX_GB: Env.schema.number.optional(),

start/routes.ts

@@ -63,6 +63,13 @@ router
       router
         .group(() => {
           router.post('lease', [controllers.Node, 'lease'])
+          // Group leases (REBUILD.md, "Voyager 3: window indexing", 1 and 5): the window worker's.
+          router.post('groups/lease', [controllers.Node, 'leaseGroup'])
+          router.post('groups/:id/next', [controllers.Node, 'nextRange'])
+          router.post('groups/:id/heartbeat', [controllers.Node, 'groupHeartbeat'])
+          router.post('groups/:id/release', [controllers.Node, 'releaseGroup'])
+          router.put('groups/:id/snapshot', [controllers.Node, 'putSnapshot'])
+          router.get('groups/:id/snapshot', [controllers.Node, 'getSnapshot'])
           router.get('ranges/:id/stream', [controllers.Node, 'stream'])
           router.post('ranges/:id/batches', [controllers.Node, 'batch'])
           router.post('ranges/:id/heartbeat', [controllers.Node, 'heartbeat'])

tests/bootstrap.ts

@@ -8,6 +8,7 @@ import testUtils from '@adonisjs/core/services/test_utils'
 import { authApiClient } from '@adonisjs/auth/plugins/api_client'
 import { sessionApiClient } from '@adonisjs/session/plugins/api_client'
 import type { Registry } from '../.adonisjs/client/registry/schema.d.ts'
+import CatalogRefs from '#services/catalog_refs'
 
 /**
  * This file is imported by the "bin/test.ts" entrypoint file
@@ -51,6 +52,8 @@ export const runnerHooks: Required<Pick<Config, 'setup' | 'teardown'>> = {
  */
 export const configureSuite: Config['configureSuite'] = (suite) => {
   if (['browser', 'functional', 'e2e'].includes(suite.name)) {
+    // Each test's transaction is rolled back: the poster and group-set ids cached meanwhile are gone with it.
+    suite.onGroup((group) => group.each.setup(() => CatalogRefs.forget()))
     return suite.setup(() => testUtils.httpServer().start())
   }
 }

tests/e2e/edge_node.spec.ts

@@ -0,0 +1,238 @@
+import { test } from '@japa/runner'
+import db from '@adonisjs/lucid/services/db'
+import testUtils from '@adonisjs/core/services/test_utils'
+import { mkdtemp, rm } from 'node:fs/promises'
+import { tmpdir } from 'node:os'
+import { join } from 'node:path'
+import Node from '#models/node'
+import Spool from '#services/spool'
+import Corpus from '#services/corpus'
+import GroupLeases from '#services/group_leases'
+import coordinatorConfig from '#config/coordinator'
+import { GROUP, overviewChunk, runEdge, startEdge, waitFor } from '#tests/e2e/helpers/edge'
+
+/**
+ * The Rust edge worker against the real node API, database, spool and corpus (REBUILD.md, "Voyager 3: window
+ * indexing", 1 and 5): what broke in production on 2026-10-27 (LOG.md (118)) — a stream waiting on a chunk the
+ * corpus fill stored, a failed stream, a restart with and without the snapshot, a takeover.
+ */
+test.group('Edge worker end to end', (group) => {
+  const saved = {
+    spool: coordinatorConfig.spool.dir,
+    corpus: coordinatorConfig.corpus.dir,
+    snapshots: coordinatorConfig.snapshotDir,
+    chunk: coordinatorConfig.chunkArticles,
+    wait: coordinatorConfig.streamChunkWaitMs,
+  }
+  let dir = ''
+  group.each.setup(async () => {
+    dir = await mkdtemp(join(tmpdir(), 'e2e-'))
+    coordinatorConfig.spool.dir = join(dir, 'spool')
+    coordinatorConfig.corpus.dir = join(dir, 'corpus')
+    coordinatorConfig.snapshotDir = join(dir, 'snapshots')
+    coordinatorConfig.chunkArticles = 100
+    return testUtils.db().truncate()
+  })
+  group.each.teardown(async () => {
+    Object.assign(coordinatorConfig.spool, { dir: saved.spool })
+    coordinatorConfig.corpus.dir = saved.corpus
+    coordinatorConfig.snapshotDir = saved.snapshots
+    coordinatorConfig.chunkArticles = saved.chunk
+    coordinatorConfig.streamChunkWaitMs = saved.wait
+    await rm(dir, { recursive: true, force: true })
+  })
+
+  /** A group with one range of three chunks and a node with a token. */
+  async function setup(opts: { capture?: boolean } = {}) {
+    const [g] = await db.table('groups').insert({ name: GROUP, priority: 1 }).returning('id')
+    const [r] = await db
+      .table('ranges')
+      .insert({
+        group_id: g.id,
+        start_id: 1,
+        end_id: 300,
+        next_id: 1,
+        priority: 0,
+        capture: opts.capture ?? false,
+      })
+      .returning('id')
+    const node = await Node.create({ name: 'e2e-node' })
+    const token = (await Node.accessTokens.create(node)).value!.release()
+    return { groupId: Number(g.id), rangeId: Number(r.id), nodeId: Number(node.id), token }
+  }
+  const range = async (id: number) => (await db.from('ranges').where('id', id).firstOrFail()) as any
+  const lease = async (groupId: number) =>
+    (await db
+      .from('group_leases')
+      .where('group_id', groupId)
+      .where('lane', 'history')
+      .first()) as any
+  /** The release whose first article is `at` (a release's name isn't the file's: the identity strips it). */
+  const releaseAt = async (at: number) =>
+    (await db.from('releases').where('first_article', at).first()) as any
+  const names = async () =>
+    JSON.stringify(
+      await db
+        .from('releases')
+        .select('name', 'first_article', 'last_article', 'segments', 'complete')
+        .orderBy('first_article')
+    )
+  const has = (lines: string[], text: string) => lines.some((l) => l.includes(text))
+
+  test('a range streamed from the spool and the corpus, with a chunk the corpus fill stores while the stream waits', async ({
+    assert,
+  }) => {
+    const { groupId, rangeId, token } = await setup({ capture: true })
+    await Spool.write(rangeId, { from: 1, to: 100 }, overviewChunk(1, 100))
+    await Spool.write(rangeId, { from: 201, to: 300 }, overviewChunk(201, 300))
+    const run = await startEdge(token)
+    // The worker holds the group and streams: chunk 1 goes out, chunk 2 is waited for — the corpus fill stores it.
+    await waitFor(
+      'the group to be held',
+      async () =>
+        (await lease(groupId))?.range_id !== null && (await lease(groupId))?.range_id !== undefined
+    )
+    await new Promise((r) => setTimeout(r, 700))
+    await Corpus.write(
+      { group: GROUP, start: 1, end: 300 },
+      { from: 101, to: 200 },
+      overviewChunk(101, 200),
+      3
+    )
+    const { code, lines } = { ...(await run.exit), lines: run.lines }
+    assert.equal(code, 0, lines.join('\n'))
+    assert.isTrue(has(lines, 'caught up after 1 ranges'), lines.join('\n'))
+    assert.isTrue(has(lines, 'snapshot at 301'), lines.join('\n'))
+    assert.include(await range(rangeId), { status: 'done' })
+    assert.equal(Number((await range(rangeId)).headers), 300)
+    const l = await lease(groupId)
+    assert.isNull(l.node_id)
+    assert.equal(Number(l.snapshot_next), 301)
+    assert.equal(Number(l.ranges_done), 1)
+    // The window closed on the early files (the frontier is a day past them); every closed release is settled.
+    const releases = await db
+      .from('releases')
+      .select('name', 'settled_at', 'first_article', 'last_article', 'segments')
+    assert.isAtLeast(releases.length, 10)
+    assert.isTrue(releases.every((r) => r.settled_at !== null))
+    const spanning = releases.find((r) => Number(r.first_article) === 100)
+    assert.isDefined(spanning, await names())
+    assert.deepEqual(
+      [Number(spanning.first_article), Number(spanning.last_article), Number(spanning.segments)],
+      [100, 101, 2],
+      'the file over the chunk boundary came through the carried window'
+    )
+  })
+
+  test('a stream that fails: the batch so far is posted, the window snapshotted, the group given back; the next run restores it', async ({
+    assert,
+  }) => {
+    coordinatorConfig.streamChunkWaitMs = 1500
+    const { groupId, rangeId, token } = await setup()
+    await Spool.write(rangeId, { from: 1, to: 100 }, overviewChunk(1, 100))
+    const first = await runEdge(token)
+    assert.equal(first.code, 0, first.lines.join('\n'))
+    assert.isTrue(has(first.lines, 'the group goes back'), first.lines.join('\n'))
+    assert.isTrue(has(first.lines, 'snapshot at 101'), first.lines.join('\n'))
+    let r = await range(rangeId)
+    assert.include(r, { status: 'planned', node_id: null })
+    assert.equal(
+      Number(r.next_id),
+      101,
+      'the batch for chunk 1 was stored before the group went back'
+    )
+    let l = await lease(groupId)
+    assert.isNull(l.node_id)
+    assert.equal(Number(l.snapshot_next), 101)
+    const before = await db.from('releases').count('* as n').first()
+    assert.isAtLeast(Number(before!.n), 10)
+    assert.isNull(await releaseAt(100), 'the file over the boundary is still open in the window')
+
+    // The next run: the window restored, the rest of the range streamed, the open file completed.
+    coordinatorConfig.streamChunkWaitMs = saved.wait
+    await Spool.write(rangeId, { from: 101, to: 200 }, overviewChunk(101, 200))
+    await Spool.write(rangeId, { from: 201, to: 300 }, overviewChunk(201, 300))
+    const second = await runEdge(token)
+    assert.equal(second.code, 0, second.lines.join('\n'))
+    assert.isTrue(
+      has(second.lines, 'window restored from the snapshot at 101'),
+      second.lines.join('\n')
+    )
+    r = await range(rangeId)
+    assert.equal(r.status, 'done')
+    assert.equal(Number(r.headers), 300)
+    l = await lease(groupId)
+    assert.equal(Number(l.snapshot_next), 301)
+    const spanning = await releaseAt(100)
+    assert.isNotNull(spanning, await names())
+    assert.deepEqual(
+      [Number(spanning.first_article), Number(spanning.last_article), Number(spanning.segments)],
+      [100, 101, 2]
+    )
+  })
+
+  test('a restart whose snapshot file is gone: the window starts empty and the file over the seam ends incomplete', async ({
+    assert,
+  }) => {
+    coordinatorConfig.streamChunkWaitMs = 1500
+    const { groupId, rangeId, token } = await setup()
+    await Spool.write(rangeId, { from: 1, to: 100 }, overviewChunk(1, 100))
+    const first = await runEdge(token)
+    assert.isTrue(has(first.lines, 'snapshot at 101'), first.lines.join('\n'))
+    await rm(GroupLeases.snapshotPath(groupId), { force: true })
+
+    coordinatorConfig.streamChunkWaitMs = saved.wait
+    await Spool.write(rangeId, { from: 101, to: 200 }, overviewChunk(101, 200))
+    await Spool.write(rangeId, { from: 201, to: 300 }, overviewChunk(201, 300))
+    const second = await runEdge(token)
+    assert.equal(second.code, 0, second.lines.join('\n'))
+    assert.isTrue(
+      has(second.lines, 'but there is none; the window starts empty'),
+      second.lines.join('\n')
+    )
+    assert.equal((await range(rangeId)).status, 'done')
+    // Part 1 was in the lost window: one seam. Part 2 alone is a file short of a part.
+    assert.isNull(await releaseAt(100), await names())
+    const half = await releaseAt(101)
+    assert.isNotNull(half, await names())
+    assert.deepEqual([Number(half.segments), half.complete], [1, false], await names())
+  })
+
+  test("a takeover: the stale holder's next batch is refused, its snapshot too, and the group stays the new holder's", async ({
+    assert,
+  }) => {
+    const { groupId, rangeId, token, nodeId } = await setup()
+    await Spool.write(rangeId, { from: 1, to: 100 }, overviewChunk(1, 100))
+    const run = await startEdge(token)
+    // Chunk 1 posted; the stream now waits for chunk 2.
+    await waitFor('the first batch', async () => Number((await range(rangeId)).next_id) === 101)
+    // The holder looks dead (its group and range heartbeats stale): another node takes the group over (no snapshot
+    // yet: the range goes back to its resume point).
+    const stale = new Date(Date.now() - 24 * 3600_000)
+    await db.from('group_leases').where('group_id', groupId).update({ heartbeat_at: stale })
+    await db.from('ranges').where('id', rangeId).update({ heartbeat_at: stale })
+    const other = await Node.create({ name: 'e2e-other' })
+    const taken = await GroupLeases.lease(Number(other.id))
+    assert.equal(taken?.group.id, groupId)
+    assert.isNull(taken?.snapshot)
+    assert.include(await range(rangeId), { status: 'planned', node_id: null })
+    // The old holder gets chunk 2 all the same, posts, and is refused.
+    await Spool.write(rangeId, { from: 101, to: 200 }, overviewChunk(101, 200))
+    await Spool.write(rangeId, { from: 201, to: 300 }, overviewChunk(201, 300))
+    const { code } = await run.exit
+    const lines = run.lines
+    assert.equal(code, 0, lines.join('\n'))
+    assert.isTrue(has(lines, 'snapshot failed'), lines.join('\n'))
+    assert.isTrue(
+      has(lines, 'not leased to this node') || has(lines, 'no longer this node'),
+      lines.join('\n')
+    )
+    const l = await lease(groupId)
+    assert.equal(Number(l.node_id), Number(other.id))
+    assert.isNull(l.snapshot_next)
+    const r = await range(rangeId)
+    assert.equal(r.status, 'planned')
+    assert.equal(Number(r.next_id), 101, 'nothing past the first batch was stored')
+    assert.notEqual(Number(r.node_id), nodeId)
+  })
+})

tests/e2e/helpers/edge.ts

@@ -0,0 +1,149 @@
+import { spawn } from 'node:child_process'
+import { access } from 'node:fs/promises'
+import { resolve } from 'node:path'
+import { deflateSync } from 'node:zlib'
+import env from '#start/env'
+
+export const GROUP = 'alt.binaries.e2e'
+
+/** The edge worker binary: EDGE_NODE_BIN, else the release or debug build under edge/target. */
+export async function edgeBinary(): Promise<string> {
+  const candidates = [
+    process.env.EDGE_NODE_BIN,
+    'edge/target/release/edge-node',
+    'edge/target/debug/edge-node',
+  ]
+    .filter((c): c is string => !!c)
+    .map((c) => resolve(c))
+  for (const path of candidates) {
+    try {
+      await access(path)
+      return path
+    } catch {}
+  }
+  throw new Error(
+    `No edge-node binary (looked at ${candidates.join(', ')}): build it with "cargo build --release --manifest-path edge/Cargo.toml --bin edge-node", or set EDGE_NODE_BIN`
+  )
+}
+
+/** The test server's address for the worker (tests/bootstrap.ts starts it on HOST:PORT). */
+export function coordinatorUrl(): string {
+  const host = env.get('HOST', 'localhost')
+  return `http://${host === '0.0.0.0' ? '127.0.0.1' : host}:${env.get('PORT', 3333)}`
+}
+
+export interface EdgeRun {
+  exit: Promise<{ code: number | null; signal: string | null }>
+  lines: string[]
+  kill: () => void
+}
+
+/**
+ * The real edge worker against the test server: one thread on the history lane, one group, a batch per chunk
+ * (100 articles in these tests), a snapshot after every range.
+ */
+export async function startEdge(token: string, args: string[] = []): Promise<EdgeRun> {
+  const bin = await edgeBinary()
+  // The test server first (tests/bootstrap.ts starts it for this suite): a worker that can't connect retries for
+  // minutes before it says so.
+  const probe = await fetch(`${coordinatorUrl()}/api/v1/node/lease`, { method: 'POST' }).catch(
+    (e) => e as Error
+  )
+  if (probe instanceof Error || probe.status !== 401)
+    throw new Error(
+      `The test server doesn't answer at ${coordinatorUrl()}: ${probe instanceof Error ? probe.message : `HTTP ${probe.status}`}`
+    )
+  const child = spawn(
+    bin,
+    [
+      '--threads',
+      '1',
+      '--tail-threads',
+      '0',
+      '--groups',
+      '1',
+      '--batch-articles',
+      '100',
+      '--snapshot-minutes',
+      '0',
+      ...args,
+    ],
+    {
+      env: { ...process.env, COORDINATOR_URL: coordinatorUrl(), NODE_TOKEN: token },
+      stdio: ['ignore', 'pipe', 'pipe'],
+    }
+  )
+  const lines: string[] = []
+  let rest = ''
+  const collect = (chunk: Buffer) => {
+    rest += chunk.toString('utf8')
+    const parts = rest.split('\n')
+    rest = parts.pop() ?? ''
+    lines.push(...parts)
+  }
+  child.stdout.on('data', collect)
+  child.stderr.on('data', collect)
+  const exit = new Promise<{ code: number | null; signal: string | null }>((done) => {
+    child.on('exit', (code, signal) => {
+      if (rest) lines.push(rest)
+      done({ code, signal })
+    })
+  })
+  return { exit, lines, kill: () => child.kill('SIGKILL') }
+}
+
+/** Runs the worker to its end (at most `timeoutMs`), its log lines and exit code. */
+export async function runEdge(token: string, args: string[] = [], timeoutMs = 60_000) {
+  const run = await startEdge(token, args)
+  const timer = setTimeout(() => run.kill(), timeoutMs)
+  const result = await run.exit
+  clearTimeout(timer)
+  return { ...result, lines: run.lines }
+}
+
+/** Polls `condition` every 100 ms until it holds or `timeoutMs` passed (then throws). */
+export async function waitFor(what: string, condition: () => Promise<boolean>, timeoutMs = 30_000) {
+  const until = Date.now() + timeoutMs
+  while (Date.now() < until) {
+    if (await condition()) return
+    await new Promise((r) => setTimeout(r, 100))
+  }
+  throw new Error(`Timed out waiting for ${what}`)
+}
+
+const BASE_MS = Date.UTC(2026, 9, 5, 10)
+
+/** A file name without digits (a number sequence would make an album of a poster's files): letters from `k`. */
+export function nameOf(k: number): string {
+  let s = ''
+  let n = k
+  do {
+    s = String.fromCharCode(97 + (n % 26)) + s
+    n = Math.floor(n / 26)
+  } while (n > 0)
+  return `Set.${s}.mkv`
+}
+
+/**
+ * The overview lines of articles `from`..`to`, one an hour from 2026-10-05: article 1 is a one-part file, and from 2
+ * on each even article is part 1 of a two-part file whose part 2 is the next (odd) article — so a file spans every
+ * chunk boundary at an even `to` (100 and 101, 200 and 201: the window has to carry it). Dates an hour apart make
+ * the posting frontier move a day every 24 articles, so the window closes on files as it goes.
+ */
+export function overviewLines(from: number, to: number): string[] {
+  const lines: string[] = []
+  for (let n = from; n <= to; n++) {
+    const date = new Date(BASE_MS + n * 3600_000).toUTCString().replace(' GMT', ' +0000')
+    const [name, part] =
+      n === 1 ? [nameOf(1), '(1/1)'] : n % 2 === 0 ? [nameOf(n), '(1/2)'] : [nameOf(n - 1), '(2/2)']
+    lines.push(
+      `${n}\t"${name}" yEnc ${part}\tposter <p@x>\t${date}\t<a${n}@e2e>\t\t700000\t5000\tXref: news.example ${GROUP}:${n}`
+    )
+  }
+  return lines
+}
+
+/** One XOVER chunk as the news server sends it (deflated). */
+export function overviewChunk(from: number, to: number): Buffer {
+  return deflateSync(Buffer.from(overviewLines(from, to).join('\r\n') + '\r\n'))
+}

tests/functional/batch_store.spec.ts

@@ -2,19 +2,35 @@ import { test } from '@japa/runner'
 import db from '@adonisjs/lucid/services/db'
 import testUtils from '@adonisjs/core/services/test_utils'
 import BatchStore, { StaleBatch } from '#services/batch_store'
-import { RangeIndexer } from '#lib/edge/index'
-import { decodeChunk } from '#lib/chunks/index'
+import { RangeIndexer, type EdgeBatch, type ReleaseChunk } from '#lib/edge/index'
+import { decodeChunk, encodeChunk } from '#lib/chunks/index'
 import { DEFAULT_RULES } from '#lib/rules/index'
 import type { Overview } from '#lib/nntp/index'
 
 const GROUP = 'alt.binaries.test'
 let id = 0
 
-function post(at: number, name: string, parts: number, counter: string, tool: 'ngPost' | 'plain' = 'plain'): Overview[] {
+function post(
+  at: number,
+  name: string,
+  parts: number,
+  counter: string,
+  tool: 'ngPost' | 'plain' = 'plain'
+): Overview[] {
   return Array.from({ length: parts }, (_, i) => {
     id++
     const mid = tool === 'ngPost' ? `${id.toString(16).padStart(32, '0')}@ngPost` : `art${id}@test`
-    return [at + i, { 'subject': `${counter} - "${name}" yEnc (${i + 1}/${parts})`, 'from': 'p <p@x>', 'date': 'Mon, 05 Oct 2026 10:00:00 +0000', 'message-id': `<${mid}>`, ':bytes': '700000', 'xref': `news.example ${GROUP}:${at + i}` }] as Overview
+    return [
+      at + i,
+      {
+        'subject': `${counter} - "${name}" yEnc (${i + 1}/${parts})`,
+        'from': 'p <p@x>',
+        'date': 'Mon, 05 Oct 2026 10:00:00 +0000',
+        'message-id': `<${mid}>`,
+        ':bytes': '700000',
+        'xref': `news.example ${GROUP}:${at + i}`,
+      },
+    ] as Overview
   })
 }
 
@@ -23,73 +39,224 @@ async function leasedRange() {
   const [node] = await db.table('nodes').insert({ name: 'test-node' }).returning('id')
   const [range] = await db
     .table('ranges')
-    .insert({ group_id: group.id, start_id: 1, end_id: 1_000_000, next_id: 1, status: 'leased', node_id: node.id, rule_version: 1 })
+    .insert({
+      group_id: group.id,
+      start_id: 1,
+      end_id: 1_000_000,
+      next_id: 1,
+      status: 'leased',
+      node_id: node.id,
+      rule_version: 1,
+    })
     .returning('id')
   return { rangeId: Number(range.id), nodeId: Number(node.id) }
 }
 
 function indexer() {
-  return new RangeIndexer({ group: GROUP, start: 1, end: 1_000_000, rules: { ...DEFAULT_RULES, samplePercent: 0, droppedSignatures: [], priorities: { [GROUP]: 1 } }, edgeArticles: 1000 })
+  return new RangeIndexer({
+    group: GROUP,
+    start: 1,
+    end: 1_000_000,
+    rules: {
+      ...DEFAULT_RULES,
+      samplePercent: 0,
+      droppedSignatures: [],
+      priorities: { [GROUP]: 1 },
+    },
+    edgeArticles: 1000,
+  })
 }
 
 /** One file's articles, posted `minutes` apart from `start` (an ISO time), in `group`. */
-function upload(at: number, name: string, parts: number, counter: string, start: string, group = GROUP, minutes = 1): Overview[] {
+function upload(
+  at: number,
+  name: string,
+  parts: number,
+  counter: string,
+  start: string,
+  group = GROUP,
+  minutes = 1
+): Overview[] {
   return Array.from({ length: parts }, (_, i) => {
     id++
-    const date = new Date(Date.parse(start) + i * minutes * 60_000).toUTCString().replace('GMT', '+0000')
-    return [at + i, { 'subject': `${counter} - "${name}" yEnc (${i + 1}/${parts})`, 'from': 'p <p@x>', 'date': date, 'message-id': `<up${id}@test>`, ':bytes': '700000', 'xref': `news.example ${group}:${at + i}` }] as Overview
+    const date = new Date(Date.parse(start) + i * minutes * 60_000)
+      .toUTCString()
+      .replace('GMT', '+0000')
+    return [
+      at + i,
+      {
+        'subject': `${counter} - "${name}" yEnc (${i + 1}/${parts})`,
+        'from': 'p <p@x>',
+        'date': date,
+        'message-id': `<up${id}@test>`,
+        ':bytes': '700000',
+        'xref': `news.example ${group}:${at + i}`,
+      },
+    ] as Overview
   })
 }
 
 async function rangeIn(groupName: string, start: number, nodeId?: number) {
-  const g = (await db.from('groups').where('name', groupName).first()) ?? (await db.table('groups').insert({ name: groupName, priority: groupName === GROUP ? 1 : 2 }).returning('*'))[0]
-  const node = nodeId ?? Number((await db.table('nodes').insert({ name: `n${groupName}${start}` }).returning('id'))[0].id)
-  const [range] = await db.table('ranges').insert({ group_id: g.id, start_id: start, end_id: start + 99_999, next_id: start, status: 'leased', node_id: node, rule_version: 1 }).returning('id')
+  const g =
+    (await db.from('groups').where('name', groupName).first()) ??
+    (
+      await db
+        .table('groups')
+        .insert({ name: groupName, priority: groupName === GROUP ? 1 : 2 })
+        .returning('*')
+    )[0]
+  const node =
+    nodeId ??
+    Number(
+      (
+        await db
+          .table('nodes')
+          .insert({ name: `n${groupName}${start}` })
+          .returning('id')
+      )[0].id
+    )
+  const [range] = await db
+    .table('ranges')
+    .insert({
+      group_id: g.id,
+      start_id: start,
+      end_id: start + 99_999,
+      next_id: start,
+      status: 'leased',
+      node_id: node,
+      rule_version: 1,
+    })
+    .returning('id')
   return { rangeId: Number(range.id), nodeId: node }
 }
 
+function batchOf(groupName: string, start: number, overviews: Overview[]): EdgeBatch {
+  const ix = new RangeIndexer({
+    group: groupName,
+    start,
+    end: start + 99_999,
+    rules: {
+      ...DEFAULT_RULES,
+      samplePercent: 0,
+      droppedSignatures: [],
+      priorities: { [GROUP]: 1, 'alt.binaries.other': 2 },
+    },
+    edgeArticles: 1,
+  })
+  ix.add(overviews, start + 100_000)
+  ix.finish()
+  return ix.take()
+}
+
 async function storeIn(groupName: string, start: number, overviews: Overview[]) {
   const { rangeId, nodeId } = await rangeIn(groupName, start)
-  const ix = new RangeIndexer({ group: groupName, start, end: start + 99_999, rules: { ...DEFAULT_RULES, samplePercent: 0, droppedSignatures: [], priorities: { [GROUP]: 1, 'alt.binaries.other': 2 } }, edgeArticles: 1 })
-  ix.add(overviews, start + 100_000)
-  ix.finish()
-  await BatchStore.store(rangeId, nodeId, ix.take())
+  return BatchStore.store(rangeId, nodeId, batchOf(groupName, start, overviews))
 }
 
-const releasesOf = () => db.from('releases').orderBy('id').select('release_key', 'identity_key', 'files', 'posted_at', 'last_posted_at')
+const releasesOf = () =>
+  db
+    .from('releases')
+    .orderBy('id')
+    .select(
+      db.raw(
+        `encode(release_key, 'hex') AS release_key, encode(identity_key, 'hex') AS identity_key, files, chunks, posted_at, last_posted_at, settled_at`
+      )
+    )
 
 test.group('BatchStore', (group) => {
   group.each.setup(() => testUtils.db().withGlobalTransaction())
 
-  test('stores releases, their chunks, probe candidates, rule counts and fragments, and moves the resume point', async ({ assert }) => {
+  test('stores releases by id references, their texts, chunks, probe candidates and rule counts, and moves the resume point', async ({
+    assert,
+  }) => {
     const { rangeId, nodeId } = await leasedRange()
     const ix = indexer()
-    ix.add([
-      ...post(100_000, 'Some.Show.S01E01.1080p.WEB.part1.rar', 2, 'Some.Show.S01E01 [1/2]'),
-      ...post(100_010, 'Some.Show.S01E01.1080p.WEB.par2', 1, 'Some.Show.S01E01 [2/2]'),
-      ...post(100_020, 'Hs7dKq2LmZ9xWv3RtYp1Qa.part01.rar', 3, '[1/1]', 'ngPost'),
-      ...post(10, 'Edge.Cut.2020.part1.rar', 3, 'Edge.Cut [1/1]').slice(1),
-    ], 500_000)
+    ix.add(
+      [
+        ...post(100_000, 'Some.Show.S01E01.1080p.WEB.part1.rar', 2, 'Some.Show.S01E01 [1/2]'),
+        ...post(100_010, 'Some.Show.S01E01.1080p.WEB.par2', 1, 'Some.Show.S01E01 [2/2]'),
+        ...post(100_020, 'Hs7dKq2LmZ9xWv3RtYp1Qa.part01.rar', 3, '[1/1]', 'ngPost'),
+      ],
+      500_000
+    )
     const first = ix.take()
     const stored = await BatchStore.store(rangeId, nodeId, first)
-    assert.include(stored, { releases: 1, newReleases: 1, candidates: 2, fragments: 0 })
+    assert.deepEqual(stored, {
+      releases: 1,
+      newReleases: 1,
+      finalized: 0,
+      candidates: 2,
+      pending: 0,
+    })
 
     const release = await db.from('releases').first()
-    assert.include(release, { files: 2, segments: '3', chunks: 1, readable: true, has_par2: true, file_total: 2 })
+    assert.include(release, {
+      files: 2,
+      segments: '3',
+      chunks: 1,
+      readable: true,
+      has_par2: true,
+      file_total: 2,
+      complete: true,
+      edge: false,
+      source: 1,
+    })
+    assert.isNotNull(release.settled_at)
+    // The poster and the groups by id, the keys 20 bytes, the name searchable.
+    assert.equal(
+      (await db.from('posters').where('id', release.poster_id).firstOrFail()).poster,
+      'p <p@x>'
+    )
+    assert.deepEqual(
+      (await db.from('group_sets').where('id', release.group_set_id).firstOrFail()).groups,
+      [GROUP]
+    )
+    assert.lengthOf(release.release_key, 20)
+    assert.lengthOf(release.identity_key, 20)
+    assert.deepEqual(await db.from('release_texts').select('release_id', 'kind', 'text'), [
+      { release_id: release.id, kind: 1, text: 'Some Show S01E01' },
+    ])
     const chunk = await db.from('release_chunks').where('release_id', release.id).first()
-    assert.deepEqual(decodeChunk(chunk.payload).map((f) => f.name).sort(), ['Some.Show.S01E01.1080p.WEB.par2', 'Some.Show.S01E01.1080p.WEB.part1.rar'])
-    assert.deepEqual((await db.from('probe_candidates').orderBy('kind')).map((c) => c.kind), ['archive', 'par2'])
-    assert.deepEqual((await db.from('rule_counts')).map((c) => [c.rule, c.key, Number(c.articles)]), [['garbage', 'ngPost alt.binaries.test', 3]])
-    assert.equal(Number((await db.from('ranges').where('id', rangeId).first()).next_id), 500_000)
+    assert.deepEqual(
+      decodeChunk(chunk.payload)
+        .map((f) => f.name)
+        .sort(),
+      ['Some.Show.S01E01.1080p.WEB.par2', 'Some.Show.S01E01.1080p.WEB.part1.rar']
+    )
+    // Claimable at once: the release is finished.
+    assert.deepEqual(
+      (await db.from('probe_candidates').orderBy('kind')).map((c) => [
+        c.kind,
+        c.ready,
+        c.unreadable,
+      ]),
+      [
+        ['archive', true, false],
+        ['par2', true, false],
+      ]
+    )
+    // Keyed by the range's block (a range without one is a block of its own, until blocks are leased).
+    const range = await db.from('ranges').where('id', rangeId).first()
+    assert.isNotNull(range.block_id)
+    assert.equal(Number(chunk.block_id), Number(range.block_id))
+    assert.deepEqual(
+      (await db.from('rule_counts')).map((c) => [
+        Number(c.block_id),
+        c.rule,
+        c.key,
+        Number(c.articles),
+      ]),
+      [[Number(range.block_id), 'garbage', 'ngPost alt.binaries.test', 3]]
+    )
+    assert.equal(Number(range.next_id), 500_000)
 
     // The same batch again (a retry after a lost reply) is refused, and changes nothing.
     await assert.rejects(() => BatchStore.store(rangeId, nodeId, first), StaleBatch)
     assert.equal((await db.from('releases').first()).files, 2)
 
-    // The end of the range: the cut-off file near the start edge is kept as a fragment.
+    // The end of the range.
     ix.finish()
-    const last = ix.take()
-    assert.include(await BatchStore.store(rangeId, nodeId, last), { fragments: 1 })
+    await BatchStore.store(rangeId, nodeId, ix.take())
     assert.equal(Number((await db.from('ranges').where('id', rangeId).first()).next_id), 1_000_001)
   })
 
@@ -97,100 +264,317 @@ test.group('BatchStore', (group) => {
     const { rangeId, nodeId } = await leasedRange()
     const ix = indexer()
     ix.add(post(100_000, 'A.Release.2020.mkv', 1, 'A [1/1]'), 200_000)
-    await assert.rejects(() => BatchStore.store(rangeId, nodeId + 1, ix.take()), /not leased to this node/)
+    await assert.rejects(
+      () => BatchStore.store(rangeId, nodeId + 1, ix.take()),
+      /not leased to this node/
+    )
   })
-  test('a re-upload is a release of its own: in one range, or a later range; a continuation joins', async ({ assert }) => {
-    // One range: the same file posted Monday and again Thursday.
+
+  test('every chunk is a release of its own: a re-upload, and a part read by another range (the block merger joins those)', async ({
+    assert,
+  }) => {
+    // One range: the same file posted Monday and again Thursday are two uploads.
     await storeIn(GROUP, 1, [
-      ...upload(10, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/2]', '2026-10-05T10:00:00Z'),
-      ...upload(20, 'Some.Show.S01E01.1080p.par2', 1, 'Some.Show.S01E01.1080p [2/2]', '2026-10-05T10:05:00Z'),
-      ...upload(500, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/2]', '2026-10-08T10:00:00Z'),
+      ...upload(
+        10,
+        'Some.Show.S01E01.1080p.mkv',
+        2,
+        'Some.Show.S01E01.1080p [1/2]',
+        '2026-10-05T10:00:00Z'
+      ),
+      ...upload(
+        20,
+        'Some.Show.S01E01.1080p.par2',
+        1,
+        'Some.Show.S01E01.1080p [2/2]',
+        '2026-10-05T10:05:00Z'
+      ),
+      ...upload(
+        500,
+        'Some.Show.S01E01.1080p.mkv',
+        2,
+        'Some.Show.S01E01.1080p [1/2]',
+        '2026-10-08T10:00:00Z'
+      ),
     ])
     let rows = await releasesOf()
-    assert.lengthOf(rows, 2)
-    assert.equal(rows[1].release_key, `${rows[0].release_key}:2`)
-    assert.deepEqual(rows.map((r) => [r.identity_key, r.files]), [[rows[0].release_key, 2], [rows[0].release_key, 1]])
+    assert.deepEqual(
+      rows.map((r) => r.files),
+      [2, 1]
+    )
+    assert.equal(rows[0].identity_key, rows[1].identity_key)
+    assert.notEqual(rows[0].release_key, rows[1].release_key)
 
-    // The next range: Thursday's par2 an hour later joins Thursday's release; a Saturday copy is a third.
-    await storeIn(GROUP, 100_001, upload(100_010, 'Some.Show.S01E01.1080p.par2', 1, 'Some.Show.S01E01.1080p [2/2]', '2026-10-08T11:00:00Z'))
-    await storeIn(GROUP, 200_001, upload(200_010, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/2]', '2026-10-10T10:00:00Z'))
+    // The next range: Thursday's par2 an hour later is written as a release of its own (its key: the identity, the
+    // group and its first article); the window keeps such parts together, and the block merger joins the edges.
+    await storeIn(
+      GROUP,
+      100_001,
+      upload(
+        100_010,
+        'Some.Show.S01E01.1080p.par2',
+        1,
+        'Some.Show.S01E01.1080p [2/2]',
+        '2026-10-08T11:00:00Z'
+      )
+    )
     rows = await releasesOf()
-    assert.deepEqual(rows.map((r) => [r.release_key.split(':')[1] ?? '1', r.files]), [['1', 2], ['2', 2], ['3', 1]])
-    const chunks = await db.from('release_chunks').count('* as n').first()
-    assert.equal(Number(chunks!.n), 4)
+    assert.deepEqual(
+      rows.map((r) => r.files),
+      [2, 1, 1]
+    )
+    assert.lengthOf(new Set(rows.map((r) => r.release_key)), 3)
+    assert.lengthOf(new Set(rows.map((r) => r.identity_key)), 1)
+    assert.equal(Number((await db.from('release_chunks').count('* as n').first())!.n), 3)
   })
 
-  test('candidates added to a settled release (a later range of its upload) are claimable at once', async ({ assert }) => {
-    await storeIn(GROUP, 1, upload(10, 'Some.Show.S01E02.1080p.mkv', 2, 'Some.Show.S01E02.1080p [1/2]', '2026-10-05T10:00:00Z'))
-    // Settled before the next range came (the settler marks only what it settles).
-    await db.from('releases').update({ settled_at: new Date() })
-    await db.from('probe_candidates').update({ ready: true })
-    await storeIn(GROUP, 100_001, upload(100_010, 'Some.Show.S01E02.1080p.par2', 1, 'Some.Show.S01E02.1080p [2/2]', '2026-10-05T10:20:00Z'))
-    const candidates = await db.from('probe_candidates').orderBy('id').select('kind', 'ready')
-    assert.deepEqual(candidates.map((c) => c.kind), ['par2'])
-    assert.isTrue(candidates.every((c) => c.ready))
-    assert.lengthOf(await releasesOf(), 1)
-  })
+  test('a provisional tail release is finished once by a later chunk of its key, which adds the files not sent before', async ({
+    assert,
+  }) => {
+    const { rangeId, nodeId } = await rangeIn(GROUP, 1)
+    const batch = batchOf(
+      GROUP,
+      1,
+      upload(
+        10,
+        'Some.Show.S01E02.1080p.mkv',
+        2,
+        'Some.Show.S01E02.1080p [1/2]',
+        '2026-10-05T10:00:00Z'
+      )
+    )
+    assert.lengthOf(batch.releases, 1)
+    const provisional: ReleaseChunk = { ...batch.releases[0], finished: false }
+    const stored = await BatchStore.store(rangeId, nodeId, {
+      ...batch,
+      next: 50_000,
+      releases: [provisional],
+    })
+    assert.include(stored, { newReleases: 1, finalized: 0 })
+    let [row] = await releasesOf()
+    assert.isNull(row.settled_at)
+    assert.include(row, { files: 1, chunks: 1 })
 
-  test('a split 7z read in pieces keeps the tail of its highest volume only; RAR gets none', async ({ assert }) => {
-    // As the live tail reads it while posted: each piece queues a tail for the last volume it saw so far.
+    // The finish: the par2 that came within the window, under the same key (the identity, the group, the first
+    // article), with its probe candidate: claimable at once, the release being finished.
+    const second = batchOf(
+      GROUP,
+      1,
+      upload(
+        30,
+        'Some.Show.S01E02.1080p.par2',
+        1,
+        'Some.Show.S01E02.1080p [2/2]',
+        '2026-10-05T10:20:00Z'
+      )
+    )
+    const [par2] = second.releases
+    const finish: ReleaseChunk = {
+      ...par2,
+      key: provisional.key,
+      upload: provisional.upload,
+      firstArticle: provisional.firstArticle,
+      finished: true,
+    }
+    const candidates = second.candidates.map((c) => ({
+      ...c,
+      releaseKey: provisional.key,
+      upload: provisional.upload,
+    }))
+    assert.deepEqual(
+      candidates.map((c) => c.kind),
+      ['par2']
+    )
+    const finished = await BatchStore.store(rangeId, nodeId, {
+      ...batch,
+      from: 50_000,
+      next: 60_000,
+      releases: [finish],
+      candidates,
+      counts: [],
+      samples: [],
+      fragments: [],
+    })
+    assert.include(finished, { releases: 1, newReleases: 0, finalized: 1, candidates: 1 })
+    ;[row] = await releasesOf()
+    assert.isNotNull(row.settled_at)
+    assert.include(row, { files: 2, chunks: 2 })
+    const release = await db.from('releases').firstOrFail()
+    assert.include(release, { complete: true, has_par2: true, last_article: '30' })
+    assert.deepEqual(await db.from('probe_candidates').select('kind', 'ready'), [
+      { kind: 'par2', ready: true },
+    ])
+    assert.deepEqual(
+      (await db.from('release_chunks').orderBy('id')).flatMap((c) =>
+        decodeChunk(c.payload).map((f) => f.name)
+      ),
+      ['Some.Show.S01E02.1080p.mkv', 'Some.Show.S01E02.1080p.par2']
+    )
+
+    // Sent once more, finished already: left as it is.
+    const again = await BatchStore.store(rangeId, nodeId, {
+      ...batch,
+      from: 60_000,
+      next: 70_000,
+      releases: [{ ...finish, payload: encodeChunk([]), files: 0, segments: 0, bytes: 0 }],
+      candidates: [],
+      counts: [],
+      samples: [],
+      fragments: [],
+    })
+    assert.include(again, { releases: 1, newReleases: 0, finalized: 0 })
+    assert.include((await releasesOf())[0], { files: 2, chunks: 2 })
+    assert.equal(Number((await db.from('release_chunks').count('* as n').first())!.n), 2)
+  })
+
+  test('a split 7z keeps the tail of its highest volume only; RAR gets none', async ({
+    assert,
+  }) => {
     await storeIn(GROUP, 1, [
-      ...upload(10, 'TsIgNBewuFbuwWy.7z.001', 2, '[1/4]', '2026-10-23T08:24:00Z'),
-      ...upload(20, 'TsIgNBewuFbuwWy.7z.002', 2, '[2/4]', '2026-10-23T08:26:00Z'),
+      ...upload(10, 'Some.Show.S01E04.7z.001', 2, '[1/4]', '2026-10-23T08:24:00Z'),
+      ...upload(20, 'Some.Show.S01E04.7z.002', 2, '[2/4]', '2026-10-23T08:26:00Z'),
+      ...upload(30, 'Some.Show.S01E04.7z.003', 2, '[3/4]', '2026-10-23T08:28:00Z'),
+      ...upload(40, 'Some.Show.S01E04.7z.004', 2, '[4/4]', '2026-10-23T08:30:00Z'),
+      ...upload(
+        50,
+        'Some.Show.S01E03.part01.rar',
+        2,
+        'Some.Show.S01E03 [1/2]',
+        '2026-10-23T09:00:00Z'
+      ),
+      ...upload(
+        60,
+        'Some.Show.S01E03.part02.rar',
+        2,
+        'Some.Show.S01E03 [2/2]',
+        '2026-10-23T09:02:00Z'
+      ),
     ])
-    await storeIn(GROUP, 100_001, upload(100_010, 'TsIgNBewuFbuwWy.7z.004', 2, '[4/4]', '2026-10-23T08:30:00Z'))
-    // A lower volume posted late: its tail goes at once.
-    await storeIn(GROUP, 200_001, upload(200_010, 'TsIgNBewuFbuwWy.7z.003', 2, '[3/4]', '2026-10-23T08:28:00Z'))
-    await storeIn(GROUP, 300_001, upload(300_010, 'Some.Show.S01E03.part01.rar', 2, 'Some.Show.S01E03 [1/2]', '2026-10-23T09:00:00Z'))
-    await storeIn(GROUP, 400_001, upload(400_010, 'Some.Show.S01E03.part02.rar', 2, 'Some.Show.S01E03 [2/2]', '2026-10-23T09:02:00Z'))
     const candidates = await db.from('probe_candidates').orderBy('file').select('kind', 'file')
     assert.deepEqual(
       candidates.map((c) => [c.kind, c.file]),
       [
         ['archive', 'Some.Show.S01E03.part01.rar'],
-        ['archive', 'TsIgNBewuFbuwWy.7z.001'],
-        ['7z-tail', 'TsIgNBewuFbuwWy.7z.004'],
+        ['archive', 'Some.Show.S01E04.7z.001'],
+        ['7z-tail', 'Some.Show.S01E04.7z.004'],
       ]
     )
   })
 
-  test('the same upload in another group (multi-posted) is a release of its own', async ({ assert }) => {
-    await storeIn(GROUP, 1, upload(10, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/1]', '2026-10-05T10:00:00Z'))
-    await storeIn('alt.binaries.other', 1, upload(10, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/1]', '2026-10-05T10:01:00Z', 'alt.binaries.other'))
+  test('the same upload in another group (multi-posted) is a release of its own', async ({
+    assert,
+  }) => {
+    await storeIn(
+      GROUP,
+      1,
+      upload(
+        10,
+        'Some.Show.S01E01.1080p.mkv',
+        2,
+        'Some.Show.S01E01.1080p [1/1]',
+        '2026-10-05T10:00:00Z'
+      )
+    )
+    await storeIn(
+      'alt.binaries.other',
+      1,
+      upload(
+        10,
+        'Some.Show.S01E01.1080p.mkv',
+        2,
+        'Some.Show.S01E01.1080p [1/1]',
+        '2026-10-05T10:01:00Z',
+        'alt.binaries.other'
+      )
+    )
     const rows = await releasesOf()
     assert.lengthOf(rows, 2)
     assert.equal(rows[0].identity_key, rows[1].identity_key)
+    assert.notEqual(rows[0].release_key, rows[1].release_key)
   })
 
-  test('an upload spread over groups (other files in another group) is one release', async ({ assert }) => {
-    await storeIn(GROUP, 1, upload(10, 'Some.Album-WEB-2026-FTD.part1.rar', 2, 'Some.Album-WEB-2026-FTD [2/3]', '2026-10-05T10:00:00Z'))
-    await storeIn('alt.binaries.other', 1, upload(10, 'Some.Album-WEB-2026-FTD.part2.rar', 2, 'Some.Album-WEB-2026-FTD [3/3]', '2026-10-05T10:00:30Z', 'alt.binaries.other'))
-    const rows = await releasesOf()
-    assert.lengthOf(rows, 1)
-    assert.equal(rows[0].files, 2)
+  test('an unnamed release goes to the pending store with its probe targets, not into the catalog', async ({
+    assert,
+  }) => {
+    const { rangeId, nodeId } = await rangeIn(GROUP, 1)
+    const batch = batchOf(GROUP, 1, [
+      ...upload(10, 'q3B9OrpKiCxUnzgtUGnp.part01.rar', 4, '[1/2]', '2026-10-05T10:00:00Z'),
+      ...upload(20, 'q3B9OrpKiCxUnzgtUGnp.par2', 1, '[2/2]', '2026-10-05T10:01:00Z'),
+    ])
+    assert.lengthOf(batch.releases, 1)
+    assert.isFalse(batch.releases[0].readable)
+    const stored = await BatchStore.store(rangeId, nodeId, batch)
+    assert.include(stored, { releases: 1, newReleases: 0, candidates: 0, pending: 1 })
+    assert.equal(Number((await db.from('releases').count('* as n').first())!.n), 0)
+    assert.equal(Number((await db.from('probe_candidates').count('* as n').first())!.n), 0)
+    assert.equal(Number((await db.from('release_chunks').count('* as n').first())!.n), 0)
+    const pending = await db.from('pending_releases').firstOrFail()
+    assert.include(pending, {
+      files: 2,
+      segments: '5',
+      partial: false,
+      finished: true,
+      tool: 'other',
+      status: 0,
+      poster: 'p <p@x>',
+    })
+    assert.lengthOf(pending.release_key, 20)
+    assert.deepEqual(
+      pending.candidates.map((c: { kind: string; file: string }) => [c.kind, c.file]).sort(),
+      [
+        ['archive', 'q3B9OrpKiCxUnzgtUGnp.part01.rar'],
+        ['par2', 'q3B9OrpKiCxUnzgtUGnp.par2'],
+      ]
+    )
+    // Sent again as a later chunk of a tail release: its counts add up while it is open, not once finished.
+    const again = await BatchStore.store(rangeId, nodeId, {
+      ...batch,
+      from: batch.next,
+      next: batch.next + 1000,
+      releases: [{ ...batch.releases[0], files: 1, segments: 5, bytes: 100 }],
+      candidates: [],
+    })
+    assert.include(again, { pending: 0 })
+    assert.include(await db.from('pending_releases').firstOrFail(), { files: 2, segments: '5' })
   })
 
-  test('a release key taken meanwhile by another group\'s writer is skipped for the next number', async ({ assert }) => {
-    await storeIn(GROUP, 1, upload(10, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/1]', '2026-10-05T10:00:00Z'))
-    const [first] = await releasesOf()
-    // As if another group's batch numbered the key at the same moment: key:2 exists, not counted for this key.
-    const [g] = await db.table('groups').insert({ name: 'alt.binaries.third', priority: 3 }).returning('id')
-    await db.table('releases').insert({ release_key: `${first.identity_key}:2`, identity_key: 'someone-else', group_id: g.id, name: 'x', source: 'subject', poster: 'p', groups: JSON.stringify([]), files: 0, segments: 0, bytes: 0, chunks: 0 })
-    await storeIn('alt.binaries.other', 1, upload(10, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/1]', '2026-10-05T10:01:00Z', 'alt.binaries.other'))
-    const keys = (await releasesOf()).filter((r) => r.identity_key === first.identity_key).map((r) => r.release_key)
-    assert.deepEqual(keys, [first.identity_key, `${first.identity_key}:3`])
-  })
-
-  test('a re-upload is numbered after the highest number, not into the holes merges left', async ({ assert }) => {
-    for (let day = 0; day < 4; day++) {
-      await storeIn(GROUP, 1 + day * 100_000, upload(10 + day * 100_000, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/1]', `2026-10-0${day * 2 + 1}T10:00:00Z`))
+  test("an album's members stay findable by their own names (release_texts kind member)", async ({
+    assert,
+  }) => {
+    const { rangeId, nodeId } = await rangeIn(GROUP, 1)
+    // Two scans the TypeScript indexer puts under their subject; the Rust edge's album of them is the same chunk
+    // with its name and source.
+    const batch = batchOf(GROUP, 1, [
+      ...upload(10, 'Scan_001.dat', 1, 'Some Scans [1/2]', '2026-10-05T10:00:00Z'),
+      ...upload(20, 'Scan_002.dat', 1, 'Some Scans [2/2]', '2026-10-05T10:01:00Z'),
+    ])
+    assert.lengthOf(batch.releases, 1)
+    const album: ReleaseChunk = {
+      ...batch.releases[0],
+      name: 'Scan_#.dat',
+      source: 'album',
+      fileTotal: null,
     }
-    const [first] = await releasesOf()
-    // Merged away: key:2 and key:3 gone, key and key:4 left.
-    await db.from('releases').whereIn('release_key', [`${first.identity_key}:2`, `${first.identity_key}:3`]).delete()
-    await storeIn(GROUP, 400_001, upload(400_010, 'Some.Show.S01E01.1080p.mkv', 2, 'Some.Show.S01E01.1080p [1/1]', '2026-10-09T10:00:00Z'))
-    const keys = (await releasesOf()).map((r) => r.release_key.split(':')[1] ?? '1')
-    assert.deepEqual(keys, ['1', '4', '5'])
+    const stored = await BatchStore.store(rangeId, nodeId, { ...batch, releases: [album] })
+    assert.include(stored, { releases: 1, newReleases: 1 })
+    const release = await db.from('releases').firstOrFail()
+    assert.include(release, {
+      name: 'Scan_#.dat',
+      source: 6,
+      files: 2,
+      file_total: null,
+      complete: null,
+    })
+    assert.deepEqual(
+      (
+        await db.from('release_texts').where('release_id', release.id).orderBy(['kind', 'text'])
+      ).map((t) => [t.kind, t.text]),
+      [
+        [1, 'Scan_#.dat'],
+        [6, 'Scan_001.dat'],
+        [6, 'Scan_002.dat'],
+      ]
+    )
   })
 
   test('a small release of sidecars only is marked for the sidecar fold', async ({ assert }) => {
@@ -199,7 +583,12 @@ test.group('BatchStore', (group) => {
       ...upload(1_000_010, 'Some.Tool.v1.zip', 1, '[01/01]', '2026-10-05T10:00:00Z'),
       ...upload(1_000_020, 'Some.Show.S01E01.part01.rar', 2, '[02/30]', '2026-10-05T10:00:00Z'),
     ])
-    const flags = new Map((await db.from('releases').select('name', 'sidecar_only')).map((r) => [r.name, r.sidecar_only]))
+    const flags = new Map(
+      (await db.from('releases').select('name', 'sidecar_only')).map((r) => [
+        r.name,
+        r.sidecar_only,
+      ])
+    )
     assert.isTrue(flags.get('Murdoch Mysteries S8D1'))
     assert.isFalse(flags.get('Some.Tool.v1'))
     // Over 1 MB: not looked at.

tests/functional/final_completeness.spec.ts

@@ -1,95 +0,0 @@
-import { test } from '@japa/runner'
-import db from '@adonisjs/lucid/services/db'
-import testUtils from '@adonisjs/core/services/test_utils'
-import BatchStore from '#services/batch_store'
-import RangeBlocks from '#services/range_blocks'
-import SearchService from '#services/search_service'
-import { RangeIndexer, postedSpan } from '#lib/edge/index'
-import { DEFAULT_RULES } from '#lib/rules/index'
-import type { Overview } from '#lib/nntp/index'
-
-const GROUP = 'alt.binaries.final'
-const at = (iso: string) => new Date(iso)
-
-test.group('Final completeness', (group) => {
-  group.each.setup(() => testUtils.db().withGlobalTransaction())
-
-  test("a batch's posting times, without the earliest and latest 1%, become the range's span", async ({ assert }) => {
-    assert.deepEqual(postedSpan([]), [null, null])
-    const dates = Array.from({ length: 200 }, (_, i) => 1_000_000 + i)
-    dates.push(5, 9_999_999_999)
-    // 202 dates: 2 trimmed at each end, so the two wild ones go.
-    assert.deepEqual(postedSpan(dates), [1_000_001, 1_000_198])
-
-    const [g] = await db.table('groups').insert({ name: GROUP, priority: 1 }).returning('id')
-    const [node] = await db.table('nodes').insert({ name: 'final-node' }).returning('id')
-    const [range] = await db
-      .table('ranges')
-      .insert({ group_id: g.id, start_id: 1, end_id: 1000, next_id: 1, status: 'leased', node_id: node.id, rule_version: 1 })
-      .returning('id')
-    const overviews = Array.from({ length: 10 }, (_, i) => {
-      const date = new Date(Date.parse('2026-01-01T10:00:00Z') + i * 60_000).toUTCString().replace('GMT', '+0000')
-      return [1 + i, { 'subject': `[1/1] - "Some.Thing.2026.rar" yEnc (${i + 1}/10)`, 'from': 'p <p@x>', 'date': date, 'message-id': `<f${i}@x>`, ':bytes': '1000', 'xref': `x ${GROUP}:${1 + i}` }] as Overview
-    })
-    const ix = new RangeIndexer({ group: GROUP, start: 1, end: 1000, rules: { ...DEFAULT_RULES, samplePercent: 0, droppedSignatures: [], priorities: { [GROUP]: 1 } } })
-    ix.add(overviews, 11)
-    await BatchStore.store(Number(range.id), Number(node.id), ix.take())
-    const stored = await db.from('ranges').where('id', range.id).firstOrFail()
-    assert.equal(new Date(stored.posted_low).toISOString(), '2026-01-01T10:00:00.000Z')
-    assert.equal(new Date(stored.posted_high).toISOString(), '2026-01-01T10:09:00.000Z')
-  })
-
-  test('a release is final once one block of done ranges covers its posting times and 6 hours either side', async ({ assert }) => {
-    const [g] = await db.table('groups').insert({ name: GROUP, priority: 1 }).returning('id')
-    const range = (start: number, status: string, low: string | null, high: string | null) =>
-      db.table('ranges').insert({ group_id: g.id, start_id: start, end_id: start + 99_999, next_id: start, status, posted_low: low && at(low), posted_high: high && at(high) })
-    await range(1, 'done', '2026-01-01T00:00:00Z', '2026-01-01T12:00:00Z')
-    await range(100_001, 'done', '2026-01-01T12:00:00Z', '2026-01-01T23:00:00Z')
-    await range(200_001, 'planned', null, null)
-    await range(300_001, 'done', '2026-01-02T10:00:00Z', '2026-01-02T20:00:00Z')
-    assert.equal(await RangeBlocks.rebuild(), 2)
-
-    let n = 0
-    const release = async (first: number, posted: string, extra: Record<string, unknown> = {}) => {
-      n++
-      const key = `fc${n}`.padEnd(40, '0')
-      const [row] = await db
-        .table('releases')
-        .insert({ release_key: key, identity_key: key, name: `Final.Test.${n}`, source: 'stem', groups: '[]', group_id: g.id, files: 3, file_total: 5, complete: false, first_article: first, last_article: first + 10, posted_at: at(posted), last_posted_at: at(posted), readable: true, settled_at: new Date(), ...extra })
-        .returning('id')
-      return Number(row.id)
-    }
-    const final = async (id: number) => {
-      const r = await db.from('releases').where('id', id).firstOrFail()
-      const { rows } = await db.rawQuery(`SELECT release_final(?, ?, ?, ?, ?, ?) AS final`, [r.group_id, r.first_article, r.last_article, r.posted_at, r.last_posted_at, r.complete])
-      return rows[0].final as boolean
-    }
-
-    // Mid-block: final either way.
-    const middle = await release(90_000, '2026-01-01T11:00:00Z')
-    // In the first range: nothing older to wait for, although 6 hours before it isn't covered.
-    const first = await release(1_000, '2026-01-01T02:00:00Z')
-    // Near the block's end, the next range unread: its missing files may be there.
-    const edge = await release(195_000, '2026-01-01T22:00:00Z')
-    // In the island after the gap: the gap is before it.
-    const island = await release(310_000, '2026-01-02T12:00:00Z')
-    assert.isTrue(await final(middle))
-    assert.isTrue(await final(first))
-    assert.isFalse(await final(edge))
-    // Complete there: every file found, nothing to wait for.
-    assert.isTrue(await final(await release(196_000, '2026-01-01T22:30:00Z', { files: 5, complete: true, name: 'Complete.Edge' })))
-    assert.isFalse(await final(island))
-
-    // Search: an incomplete release that isn't final isn't left out as incomplete; the API says which is final.
-    const shown = await SearchService.search({ q: 'Final.Test', notIncomplete: true })
-    assert.sameMembers(shown.releases.map((r) => r.id), [edge, island])
-    assert.isTrue(shown.releases.every((r) => r.final === false))
-
-    // The gap read: one block, the edge release final (the island's 12:00 + 6 h is inside 20:00).
-    await db.from('ranges').where('start_id', 200_001).update({ status: 'done', posted_low: at('2026-01-01T23:00:00Z'), posted_high: at('2026-01-02T10:00:00Z') })
-    assert.equal(await RangeBlocks.rebuild(), 1)
-    assert.isTrue(await final(edge))
-    assert.isTrue(await final(island))
-    assert.equal((await SearchService.search({ q: 'Final.Test', notIncomplete: true })).total, 0)
-  })
-})

tests/functional/fragment_merger.spec.ts

@@ -1,65 +0,0 @@
-import { test } from '@japa/runner'
-import db from '@adonisjs/lucid/services/db'
-import testUtils from '@adonisjs/core/services/test_utils'
-import BatchStore from '#services/batch_store'
-import FragmentMerger from '#services/fragment_merger'
-import { RangeIndexer } from '#lib/edge/index'
-import { decodeChunk } from '#lib/chunks/index'
-import { DEFAULT_RULES } from '#lib/rules/index'
-import type { Overview } from '#lib/nntp/index'
-
-const GROUP = 'alt.binaries.test'
-const rules = { ...DEFAULT_RULES, samplePercent: 0, droppedSignatures: [], priorities: { [GROUP]: 1 } }
-
-function parts(name: string, total: number, numbers: number[], at: number): Overview[] {
-  return numbers.map((n, i) => [at + i, {
-    'subject': `Big.Upload.2020 [1/1] - "${name}" yEnc (${n}/${total})`, 'from': 'p <p@x>', 'date': 'Mon, 05 Oct 2026 10:00:00 +0000',
-    'message-id': `<${name}-${n}@test>`, ':bytes': '700000', 'xref': `news.example ${GROUP}:${at + i}`,
-  }] as Overview)
-}
-
-/** Indexes one range's overviews and stores them as the range's only batch; the range ends done. */
-async function indexRange(groupId: number, nodeId: number, start: number, end: number, overviews: Overview[]) {
-  const [r] = await db.table('ranges').insert({ group_id: groupId, start_id: start, end_id: end, next_id: start, status: 'leased', node_id: nodeId }).returning('id')
-  const indexer = new RangeIndexer({ group: GROUP, start, end, rules, edgeArticles: 100 })
-  indexer.add(overviews, end + 1)
-  indexer.finish()
-  await BatchStore.store(Number(r.id), nodeId, indexer.take())
-  await db.from('ranges').where('id', r.id).update({ status: 'done' })
-  return Number(r.id)
-}
-
-test.group('FragmentMerger', (group) => {
-  group.each.setup(() => testUtils.db().withGlobalTransaction())
-
-  test('a file cut by a range edge is completed from both sides once both ranges are done', async ({ assert }) => {
-    const [g] = await db.table('groups').insert({ name: GROUP, priority: 1, low: 1 }).returning('id')
-    const [n] = await db.table('nodes').insert({ name: 'merge-node' }).returning('id')
-    // Parts 1–2 at the end of the first range (still open there), 3–4 at the start of the next.
-    await indexRange(g.id, n.id, 1, 1000, parts('big.part1.rar', 4, [1, 2], 990))
-    assert.equal((await db.from('fragments').count('* as c'))[0].c, '1')
-    assert.include(await FragmentMerger.run(), { completed: 0, incomplete: 0, waiting: 1 })
-
-    await indexRange(g.id, n.id, 1001, 2000, parts('big.part1.rar', 4, [3, 4], 1001))
-    assert.include(await FragmentMerger.run(), { completed: 1, incomplete: 0 })
-    assert.equal((await db.from('fragments').count('* as c'))[0].c, '0')
-    const release = await db.from('releases').first()
-    assert.include(release, { files: 1, chunks: 1 })
-    const chunk = await db.from('release_chunks').where('release_id', release.id).first()
-    assert.deepEqual(decodeChunk(chunk.payload)[0].segments.map((s) => s[0]), [1, 2, 3, 4])
-    assert.deepEqual((await db.from('probe_candidates')).map((c) => c.kind), ['archive'])
-  })
-
-  test('a file still missing parts when all its neighbours are done is dropped', async ({ assert }) => {
-    const [g] = await db.table('groups').insert({ name: GROUP, priority: 1, low: 1 }).returning('id')
-    const [n] = await db.table('nodes').insert({ name: 'merge-node' }).returning('id')
-    await indexRange(g.id, n.id, 1, 1000, parts('gap.part1.rar', 5, [1, 2], 990))
-    await indexRange(g.id, n.id, 1001, 2000, parts('gap.part1.rar', 5, [4, 5], 1001))
-    // Its later neighbour isn't there yet: the missing part could still come.
-    assert.include(await FragmentMerger.run(), { completed: 0, incomplete: 0, waiting: 1 })
-    await indexRange(g.id, n.id, 2001, 3000, [])
-    assert.include(await FragmentMerger.run(), { completed: 0, incomplete: 1 })
-    assert.equal((await db.from('fragments').count('* as c'))[0].c, '0')
-    assert.lengthOf(await db.from('releases'), 0)
-  })
-})

tests/functional/group_leases.spec.ts

@@ -0,0 +1,223 @@
+import { test } from '@japa/runner'
+import db from '@adonisjs/lucid/services/db'
+import testUtils from '@adonisjs/core/services/test_utils'
+import { mkdtemp, rm } from 'node:fs/promises'
+import { tmpdir } from 'node:os'
+import { join } from 'node:path'
+import coordinatorConfig from '#config/coordinator'
+import GroupLeases from '#services/group_leases'
+import Leases, { LeaseLost } from '#services/leases'
+import { Readable } from 'node:stream'
+
+async function node(name: string) {
+  return Number((await db.table('nodes').insert({ name }).returning('id'))[0].id)
+}
+
+async function group(name: string, priority: number) {
+  return Number((await db.table('groups').insert({ name, priority }).returning('id'))[0].id)
+}
+
+async function range(groupId: number, start: number, priority: number, extra: Record<string, unknown> = {}) {
+  const [row] = await db
+    .table('ranges')
+    .insert({ group_id: groupId, start_id: start, end_id: start + 99, next_id: start, priority, ...extra })
+    .returning('id')
+  return Number(row.id)
+}
+
+const rangeRow = async (id: number) => (await db.from('ranges').where('id', id).firstOrFail()) as any
+
+test.group('Group leases', (g) => {
+  let dir = ''
+  let snapshotDir = ''
+  g.each.setup(async () => {
+    const rollback = await testUtils.db().withGlobalTransaction()
+    dir = await mkdtemp(join(tmpdir(), 'snapshots-'))
+    snapshotDir = coordinatorConfig.snapshotDir
+    coordinatorConfig.snapshotDir = dir
+    return rollback
+  })
+  g.each.teardown(async () => {
+    coordinatorConfig.snapshotDir = snapshotDir
+    await rm(dir, { recursive: true, force: true })
+  })
+
+  test('a lease takes the group with the most urgent planned range of its lane; its ranges come in article order; caught up it gives nothing', async ({
+    assert,
+  }) => {
+    const a = await group('alt.binaries.gl.a', 1)
+    const b = await group('alt.binaries.gl.b', 2)
+    // Group a: history at priority 0 and 1; group b: a tail range (priority -3) and history.
+    const a1 = await range(a, 1, 0)
+    const a2 = await range(a, 101, 0)
+    const a3 = await range(a, 201, 1)
+    const b2 = await range(b, 101, 0)
+    const b1 = await range(b, 1, -3)
+    const n1 = await node('gl-1')
+    const n2 = await node('gl-2')
+
+    // The history lane: both groups have priority 0 work; a has the most planned articles (300 to 100).
+    const first = await GroupLeases.lease(n1)
+    assert.equal(first?.group.id, a, 'the most planned articles first')
+    assert.equal(first?.lane, 'history')
+    assert.isNull(first?.snapshot)
+    // Held: the next node gets the other group; a third gets nothing.
+    const second = await GroupLeases.lease(n2)
+    assert.equal(second?.group.id, b)
+    assert.isNull(await GroupLeases.lease(await node('gl-3')))
+
+    // Ranges in article order, whatever their priority; the tail range isn't the history lane's.
+    assert.deepEqual((await GroupLeases.next(b, n2))?.range, { id: b2, start: 101, end: 200, next: 101 })
+    assert.include(await rangeRow(b2), { status: 'leased', node_id: n2 })
+    // Asked again before finishing: the same range.
+    assert.equal((await GroupLeases.next(b, n2))?.range.id, b2)
+    // A chunk of b2 that doesn't inflate, reported after the batch before it: the range is asked for again and
+    // comes back at its resume point (the chunk fetched plain next time).
+    await db.from('ranges').where('id', b2).update({ next_id: 151 })
+    await Leases.corrupt(b2, n2, 151)
+    assert.include(await rangeRow(b2), { status: 'planned', node_id: null })
+    assert.deepEqual((await GroupLeases.next(b, n2))?.range, { id: b2, start: 101, end: 200, next: 151 })
+    assert.deepEqual((await rangeRow(b2)).plain_chunks.map(Number), [151])
+    await db.from('ranges').where('id', b2).update({ status: 'done', next_id: 201 })
+    assert.isNull(await GroupLeases.next(b, n2), 'caught up: the tail range is the other lane\'s')
+    assert.include(await rangeRow(b1), { status: 'planned' })
+
+    // The tail lane: group b's tail, held by a third node while n2 holds b's history.
+    const n3 = await node('gl-3b')
+    const tail = await GroupLeases.lease(n3, 'tail')
+    assert.equal(tail?.group.id, b)
+    assert.equal(tail?.lane, 'tail')
+    assert.isNull(await GroupLeases.lease(await node('gl-3c'), 'tail'), 'no other group has tail work')
+    assert.deepEqual((await GroupLeases.next(b, n3, 'tail'))?.range, { id: b1, start: 1, end: 100, next: 1 })
+    // The history lane is n2's.
+    await assert.rejects(() => GroupLeases.next(b, n3), LeaseLost)
+    await db.from('ranges').where('id', b1).update({ status: 'done', next_id: 101 })
+    assert.isNull(await GroupLeases.next(b, n3, 'tail'), 'the tail lane is caught up')
+    // Each lane its own snapshot.
+    await GroupLeases.putSnapshot(b, n3, 101, Readable.from([Buffer.from('tail at 101')]), 'tail')
+    await GroupLeases.putSnapshot(b, n2, 201, Readable.from([Buffer.from('history at 201')]))
+    assert.equal((await GroupLeases.getSnapshot(b, n3, 'tail'))?.next, 101)
+    assert.equal((await GroupLeases.getSnapshot(b, n2))?.next, 201)
+    await GroupLeases.release(b, n3, 'tail')
+    assert.include(await GroupLeases.held(b, n2), { node_id: n2 })
+    await assert.rejects(() => GroupLeases.held(b, n3, 'tail'), LeaseLost)
+
+    assert.deepEqual([a1, a2, a3].length, 3)
+    await assert.rejects(() => GroupLeases.next(a, n2), LeaseLost)
+  })
+
+  test('the order mixes the group with the most planned articles and the longest unheld one', async ({ assert }) => {
+    // Four groups with priority 0 work: c was held before (5000 articles left), a, b and d never (100, 1000, 50).
+    const a = await group('alt.binaries.gl.o.a', 1)
+    const b = await group('alt.binaries.gl.o.b', 2)
+    const c = await group('alt.binaries.gl.o.c', 3)
+    const d = await group('alt.binaries.gl.o.d', 4)
+    await range(a, 1, 0)
+    for (let i = 0; i < 10; i++) await range(b, 1 + 100 * i, 0)
+    for (let i = 0; i < 50; i++) await range(c, 1 + 100 * i, 0)
+    await db.table('ranges').insert({ group_id: d, start_id: 1, end_id: 50, next_id: 1, priority: 0 })
+    await db.table('group_leases').insert({ group_id: c, lane: 'history', node_id: null, heartbeat_at: new Date(Date.now() - 3600_000) })
+    // By articles c, b, a, d; by age a, b, d (never held, by group priority), c. The leases alternate which
+    // ranking goes first: c, a, b, d (or a, c, b, d when the sequence stands on the other parity).
+    await db.rawQuery(`SELECT setval('group_lease_turns', 2)`)
+    const order = []
+    for (const name of ['o1', 'o2', 'o3', 'o4']) order.push((await GroupLeases.lease(await node(name)))?.group.id)
+    assert.deepEqual(order, [c, a, b, d])
+  })
+
+  test('a stale holder is taken over and the ranges after the snapshot are read again; a release keeps the resume point', async ({
+    assert,
+  }) => {
+    const a = await group('alt.binaries.gl.c', 1)
+    const r1 = await range(a, 1, 0)
+    const r2 = await range(a, 101, 0)
+    const r3 = await range(a, 201, 0)
+    const r4 = await range(a, 301, 0)
+    const n1 = await node('gl-4')
+    const n2 = await node('gl-5')
+    assert.equal((await GroupLeases.lease(n1))?.group.id, a)
+    // The worker did r1 and r2, snapshotted at 201 (r3's start), then got halfway through r3 and went quiet.
+    await db.from('ranges').whereIn('id', [r1, r2]).update({ status: 'done', next_id: db.raw('end_id + 1') })
+    await GroupLeases.putSnapshot(a, n1, 201, Readable.from([Buffer.from('window at 201')]))
+    await GroupLeases.next(a, n1)
+    await db.from('ranges').where('id', r3).update({ next_id: 251 })
+    await db.from('group_leases').where('group_id', a).update({ heartbeat_at: new Date(Date.now() - 24 * 3600_000) })
+
+    const lease = await GroupLeases.lease(n2)
+    assert.equal(lease?.group.id, a)
+    assert.deepEqual(lease?.snapshot, { next: 201 })
+    // r1 and r2 stay done (before the snapshot); r3 is read again from the snapshot's article, r4 untouched.
+    assert.include(await rangeRow(r1), { status: 'done' })
+    assert.include(await rangeRow(r2), { status: 'done' })
+    assert.include(await rangeRow(r3), { status: 'planned', node_id: null })
+    assert.equal(Number((await rangeRow(r3)).next_id), 201)
+    assert.include(await rangeRow(r4), { status: 'planned' })
+    const snapshot = await GroupLeases.getSnapshot(a, n2)
+    assert.equal(snapshot?.next, 201)
+    assert.equal(snapshot?.size, 'window at 201'.length)
+    await assert.rejects(() => GroupLeases.next(a, n1), LeaseLost)
+
+    // The new holder works r3, snapshots at 301 and gives the group back: r3 done, r4 waits at its start.
+    assert.equal((await GroupLeases.next(a, n2))?.range.id, r3)
+    await db.from('ranges').where('id', r3).update({ status: 'done', next_id: 301 })
+    await GroupLeases.putSnapshot(a, n2, 301, Readable.from([Buffer.from('window at 301')]))
+    assert.equal((await GroupLeases.next(a, n2))?.range.id, r4)
+    await db.from('ranges').where('id', r4).update({ next_id: 301 })
+    await GroupLeases.release(a, n2)
+    assert.include(await rangeRow(r4), { status: 'planned', node_id: null })
+    assert.isNull((await db.from('group_leases').where('group_id', a).firstOrFail()).node_id)
+    // Leased again at once (free, not stale): the snapshot at 301 covers everything done.
+    const again = await GroupLeases.lease(n1)
+    assert.deepEqual(again?.snapshot, { next: 301 })
+    assert.include(await rangeRow(r3), { status: 'done' })
+    assert.equal((await GroupLeases.next(a, n1))?.range.id, r4)
+  })
+
+  test('capture only: the history ranges sharing a window priority are neither leased nor handed out', async ({
+    assert,
+  }) => {
+    // Group e: history at priority 0 only (the planner's oldest bucket); group f: a 2016 window range (capture)
+    // between two history ranges at 0.
+    const e = await group('alt.binaries.gl.e', 1)
+    await range(e, 1, 0)
+    const f = await group('alt.binaries.gl.f', 2)
+    await range(f, 1, 0)
+    const f2 = await range(f, 101, 0, { capture: true })
+    await range(f, 201, 0)
+    const n1 = await node('gl-8')
+    const was = coordinatorConfig.leaseCaptureOnly
+    coordinatorConfig.leaseCaptureOnly = true
+    try {
+      assert.equal((await GroupLeases.lease(n1))?.group.id, f, 'the group with a capture range, not the older one')
+      assert.equal((await GroupLeases.next(f, n1))?.range.id, f2)
+      await db.from('ranges').where('id', f2).update({ status: 'done', next_id: 201 })
+      assert.isNull(await GroupLeases.next(f, n1), 'the history range after the window is not handed out')
+      assert.isNull(await GroupLeases.lease(await node('gl-9')))
+    } finally {
+      coordinatorConfig.leaseCaptureOnly = was
+    }
+    assert.equal((await GroupLeases.lease(await node('gl-10')))?.group.id, e, 'any range again without the flag')
+  })
+
+  test('a snapshot older than a done range makes that range be read again from its start', async ({ assert }) => {
+    const a = await group('alt.binaries.gl.d', 1)
+    const r1 = await range(a, 1, 0)
+    const r2 = await range(a, 101, 0)
+    const n1 = await node('gl-6')
+    await GroupLeases.lease(n1)
+    // A range after the snapshot's article done long before the snapshot was taken (the old range leases did it):
+    // not the window's, it stays done.
+    const r3 = await range(a, 201, 0)
+    await db.from('ranges').where('id', r3).update({ status: 'done', next_id: 301, done_at: new Date(Date.now() - 48 * 3600_000) })
+    // The worker snapshotted at the start, finished r1, and crashed before its next snapshot: r1 is read again.
+    await GroupLeases.putSnapshot(a, n1, 1, Readable.from([Buffer.from('window at 1')]))
+    await db.from('ranges').where('id', r1).update({ status: 'done', next_id: 101, done_at: new Date() })
+    await db.from('group_leases').where('group_id', a).update({ heartbeat_at: new Date(Date.now() - 24 * 3600_000) })
+    await GroupLeases.lease(await node('gl-7'))
+    assert.equal((await rangeRow(r1)).status, 'planned')
+    assert.equal(Number((await rangeRow(r1)).next_id), 1)
+    assert.equal((await rangeRow(r2)).status, 'planned')
+    assert.equal(Number((await rangeRow(r2)).next_id), 101)
+    assert.equal((await rangeRow(r3)).status, 'done')
+  })
+})