Rebuild: a poster's constant tag no longer makes one release of all their uploads

Release 145637 held four albums: the poster puts '[KiLL.dA.FuCkIn.
EaStErRaBbIt]----' before every upload, and the text outside the quotes
(when it names a release) was the key with the poster. The same made
one release of a poster's albums under '(PowBeertje80 Posted)' (a post
id, with the same [i/M]), of masterposts' films, and of one volume of
every set in a DVD9 post.

Now, when the file's stem describes something itself (12+ letters and
digits, two words of 3+ letters) and the text outside the quotes shares
no word with it (4+ letters as part of the stem, 3-letter words and
numbers whole; boilerplate like www/net/posted not counted; no 5+ digit
per-post id), that text is taken for the poster's tag and the stem
says which release the file is. Albums of track files ('FTA … Ketchup'),
'DIE_ZEIT_35' with '35_China_dl', 'FTD769997' posts, scene names with
abbreviated file names ('grp-sm') and generic sidecars ('file_id.diz')
keep their grouping.

releases:regroup re-keys a sample of the catalog with the current rule:
on homelab's test data 145637 → 4 albums, 111566 → 17 albums of 27
files, the rest of the subject releases that change mostly lose an ad
image ('Flyer-download-2-Day.gif'); stem releases unchanged. Ported to
edge-bench-rs: check.ts still finds all captures identical (ath 2013:
16 → 76 releases on both sides).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

lib/release/identity.ts

@@ -1,5 +1,6 @@
 import { hash as digest } from 'node:crypto'
 import { looksLikeFilename } from '../headers/file_key.js'
+import { looksRandom } from './random.js'
 
 export interface ReleaseIdentity {
   key: string
@@ -106,6 +107,39 @@ function hash(...parts: (string | number | undefined)[]) {
   return digest('sha1', parts.map((p) => p ?? '').join('\n'))
 }
 
+/** A name that says what it is: 12+ letters and digits, two words of 3+ letters. */
+function describes(stem: string): boolean {
+  return stem.replace(/[^A-Za-z0-9]/g, '').length >= 12 && /[A-Za-z]{3,}[^A-Za-z0-9]+[A-Za-z]{3,}/.test(stem)
+}
+
+/** Words of a subject that say nothing about what is posted: hosts, posting boilerplate. */
+const TAG_WORDS = new Set(['www', 'http', 'https', 'com', 'net', 'org', 'info', 'the', 'and', 'presents', 'posted', 'post', 'powered', 'yenc', 'part', 'rar', 'par2', 'nzb', 'voor', 'for', 'from'])
+
+/**
+ * Whether the subject's text outside the quotes is a poster's constant tag
+ * rather than about the file: it shares no word with the file's stem. A word
+ * of 4+ letters may be part of the stem ("ketchup" in "ketchup_dance"); a
+ * 3-letter word or a number ("35" in "35_China_dl") must be a whole word;
+ * boilerplate doesn't count. A number of 5+ digits is a per-post id
+ * ("FTD 769997", "(781442)"), so never a constant tag.
+ */
+export function isPosterTag(remainder: string, stem: string): boolean {
+  const words = stem.toLowerCase().replace(/[^a-z0-9]+/g, ' ').trim()
+  const compact = words.replace(/ /g, '')
+  const whole = new Set(words.split(' '))
+  // A per-post id, also inside a word ("FTD772544").
+  if (/\d{5,}/.test(remainder)) return false
+  for (const word of remainder.toLowerCase().split(/[^a-z0-9]+/)) {
+    if (/^\d+$/.test(word)) {
+      if (word.length >= 2 && whole.has(word)) return false
+      continue
+    }
+    if (word.length < 3 || TAG_WORDS.has(word)) continue
+    if (word.length >= 4 ? compact.includes(word) : whole.has(word)) return false
+  }
+  return true
+}
+
 /**
  * Decides which release (download) a file belongs to. Returns null when the
  * file cannot be grouped from its name alone (random extensionless names in
@@ -125,7 +159,13 @@ export function releaseIdentity(input: ReleaseInput): ReleaseIdentity | null {
 
   // Text outside the quotes that names the release, e.g. "Some.Release [01/45] - "x.rar"".
   const remainder = subjectRemainder(input.subject)
-  if (namesRelease(remainder)) {
+  // Unless it's the poster's constant tag ("[KiLL.dA…]---- (01/10) "Album.rar"", "(X Posted)"),
+  // which would make one release of all their uploads: a readable file name that shares no word
+  // with it says which release the file is (LOG.md 2026-10-09 (8)).
+  // Only when the stem itself describes something ("Cephalgy.-.Weisses.Fleisch.2003", not a scene
+  // abbreviation like "grp-sm" under "Some.Movie.2008-GRP", nor "file_id.diz").
+  const tag = describes(stem) && !looksRandom(stem) && isPosterTag(remainder, stem)
+  if (!tag && namesRelease(remainder)) {
     return {
       key: hash('subject', input.poster, remainder),
       name: remainder,
@@ -135,7 +175,7 @@ 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 (fileTotal && isPostToken(remainder)) {
+  if (!tag && fileTotal && isPostToken(remainder)) {
     return { key: hash('post', input.poster, remainder, fileTotal), name: remainder, fileTotal, source: 'post' }
   }
   // The [i/M] total is informational, not part of the key: posters often