Seed batches take file ids from the insert; merge split uploads hourly

SeedService looked up every file of a batch across all files partitions
after inserting it (15% of Postgres' active time); the merge job read
every release on each GC sweep.

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

app/services/SeedService.ts

@@ -58,10 +58,17 @@ class SeedService {
     result.files = inserted.length
     result.existing = rows.length - inserted.length
 
+    // File ids for the binaries below (files queued for inspection only): new ones from the insert,
+    // existing ones looked up. Looking up every key of the batch probed all files partitions again,
+    // 15% of Postgres' active time (LOG.md 2026-10-06 (12)).
     const fileIds = new Map<string, number>()
-    for (const row of await db.from('files').whereIn('file_key', rows.map((r) => r.file_key)).select('id', 'file_key')) {
+    for (const row of inserted) if (row.file_key) fileIds.set(row.file_key, row.id)
+    const missing = files.filter((file) => !fileIds.has(file.fileKey) && ContentService.isCandidate(file.name)).map((file) => file.fileKey)
+    for (let i = 0; i < missing.length; i += 5000) {
+      for (const row of await db.from('files').whereIn('file_key', missing.slice(i, i + 5000)).select('id', 'file_key')) {
         fileIds.set(row.file_key, Number(row.id))
       }
+    }
     // A completed binary, as worker:file would have left behind, only for files queued for content
     // inspection: par2 naming (D16) needs the fingerprint inspection stores on it. For the others
     // (~85% in a.b.ath) it was only GC churn: 1.6M rows an hour inserted and deleted (LOG.md 2026-10-06 (11)).
@@ -97,9 +104,9 @@ class SeedService {
     }
 
     // As worker:file does for a new file: an SFV posted earlier may list it (D34), then inspection.
-    const rowById = new Map(rows.map((row) => [fileIds.get(row.file_key), row]))
+    const rowByKey = new Map(rows.map((row) => [row.file_key, row]))
     const added = inserted.flatMap((file) => {
-      const row = rowById.get(file.id)
+      const row = file.file_key ? rowByKey.get(file.file_key) : undefined
       return row ? [{ id: file.id, name: row.filename, poster: row.poster, date: row.date, release_id: row.release_id }] : []
     })
     await SfvService.claimAll(added)

commands/GcScheduler.ts

@@ -34,6 +34,8 @@ export default class GcScheduler extends BaseCommand {
 
   /** When release completeness was last recomputed (it runs less often than the sweep). */
   private completenessAt = 0
+  /** When split uploads were last merged. */
+  private mergedAt = 0
 
   public async run() {
     this.logger.info(
@@ -115,9 +117,13 @@ export default class GcScheduler extends BaseCommand {
       )
     }
 
-    // Uploads split per poster (CONCEPT.md D35); the releases this empties go in a later sweep.
+    // Uploads split per poster (CONCEPT.md D35); the releases this empties go in a later sweep. It reads
+    // every release and its files: hourly, not every sweep (LOG.md 2026-10-06 (12)).
+    if (this.once || Date.now() - this.mergedAt >= gcConfig.mergeIntervalMs) {
+      this.mergedAt = Date.now()
       const merged = await ReleaseMergeService.mergeSplitPosts({ dryRun: this.dryRun })
       if (merged.releases) this.logger.info(`Split uploads: ${this.verb('merge', 'merged')} ${merged.releases} releases into ${merged.names}.`)
+    }
 
     await this.rebuildIndexes()
   }

config/gc.ts

@@ -10,6 +10,9 @@ export default {
   /** How often gc:scheduler recomputes the completeness of releases that changed (LOG.md 2026-10-06 (6)). */
   releaseCompleteIntervalMs: env.get('RELEASE_COMPLETE_INTERVAL_MS', 30 * 60 * 1000),
 
+  /** How often gc:scheduler merges uploads split per poster (D35); it reads every release. */
+  mergeIntervalMs: env.get('GC_MERGE_INTERVAL_MS', 60 * 60 * 1000),
+
   /** Time per sweep for copying old file keys into file_keys until it's complete (LOG.md 2026-10-06 (7)). */
   fileKeysFillMs: env.get('FILE_KEYS_FILL_SECONDS', 90) * 1000,
 

start/env.ts

@@ -106,6 +106,7 @@ export default await Env.create(new URL('../', import.meta.url), {
   RELEASE_COMPLETE_INTERVAL_MS: Env.schema.number.optional(),
   FILE_KEYS_FILL_SECONDS: Env.schema.number.optional(),
   GC_AWAIT_INSPECTION_SECONDS: Env.schema.number.optional(),
+  GC_MERGE_INTERVAL_MS: Env.schema.number.optional(),
   ARCHIVE_WALK_MAX_SEGMENTS: Env.schema.number.optional(),
   DEFERRED_SETTLE_MS: Env.schema.number.optional(),
   DEFERRED_EXPAND_MAX_PARTS: Env.schema.number.optional(),