Par2 merge: a set held by many releases, or a merge far larger than its set, isn't one upload

Every kantodoNNN.bin test upload carries the same par2 set: 64 releases held
it nearby and became one release of 631 files for a set of 10. A set held by
more than 4 releases nearby isn't merged, nor members with more than 3x the
set's files plus 10.

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

app/services/par2_merger.ts

@@ -5,6 +5,14 @@ import { looksRandom } from '#lib/release/random'
 
 /** Releases of a set further apart than this (posted) aren't one upload: a repost is merged separately (decision d). */
 const MAX_SPREAD_MS = 48 * 3600 * 1000
+/**
+ * A set held by more releases nearby than this is one posted again and again, not one upload's par2 (every
+ * "kantodoNNN.bin" test upload carries the same set: 64 releases became one of 631 files; LOG.md 2026-10-12 (30)).
+ */
+const MAX_HOLDERS = 4
+/** Members with more files than this many times the set's, plus SLACK_FILES (its par2 volumes), are more than one upload. */
+const MAX_FILES_PER_SET_FILE = 3
+const SLACK_FILES = 10
 /** Between rounds. */
 const ROUND_MS = 60_000
 /** A round looks back this far past the previous one's start (rows committed while it ran). */
@@ -113,14 +121,19 @@ export class Par2Merger {
         releases.set(String(m.id), m)
         byFile.set(m.file_id, releases)
       }
+      const nearHolders = holders.filter(near)
+      if (nearHolders.length > MAX_HOLDERS) return 0
       const found = new Map<string, Member>()
-      for (const r of holders.filter(near)) found.set(String(r.id), r)
+      for (const r of nearHolders) found.set(String(r.id), r)
       for (const releases of byFile.values()) if (releases.size === 1) for (const [id, r] of releases) found.set(id, r)
       if (found.size < 2) return 0
       // Locked in id order (the batch store and other rounds take the same rows).
       const { rows: locked } = await trx.rawQuery(`SELECT id FROM releases WHERE id = ANY(?) ORDER BY id FOR UPDATE`, [[...found.keys()]])
       const members: Member[] = locked.map((r: { id: string }) => found.get(String(r.id))!)
       if (members.length < 2) return 0
+      const { rows: described } = await trx.rawQuery(`SELECT count(*) AS n FROM par2_files WHERE set_id = ?`, [setId])
+      const setFiles = Math.max(Number(described[0].n), 1)
+      if (members.reduce((n, m) => n + Number(m.files), 0) > MAX_FILES_PER_SET_FILE * setFiles + SLACK_FILES) return 0
       const target = members.reduce((a, b) => (Number(b.files) > Number(a.files) ? b : a))
       const others = members.filter((m) => m.id !== target.id).map((m) => Number(m.id))
       const name = members.every((m) => !m.readable) ? await this.setName(trx, setId) : undefined

tests/functional/par2_merger.spec.ts

@@ -86,6 +86,24 @@ test.group('Par2Merger', (group) => {
     assert.deepEqual(held.map((r) => Number(r.release_id)), [ids[2]])
   })
 
+  test('a set held by many releases nearby is posted again and again: not merged', async ({ assert }) => {
+    const at = new Date('2022-07-08T03:46:00Z')
+    const setId = hash('4')
+    const ids: number[] = []
+    for (let i = 1; i <= 6; i++) {
+      const key = `cc00000${i}`
+      const [row] = await db
+        .table('releases')
+        .insert({ release_key: key.padEnd(40, '0'), identity_key: key.padEnd(40, '0'), name: `kantodo${i}`, source: 'stem', groups: '[]', group_id: 1, files: 11, posted_at: at, readable: true })
+        .returning('id')
+      ids.push(Number(row.id))
+      await db.table('par2_sets').insert({ set_id: setId, release_id: row.id, files: 1 })
+    }
+    await db.table('par2_files').insert({ set_id: setId, file_id: hash('3'), hash16k: hash('2'), length: 1073741824, name: 'kantodo.bin' })
+    assert.equal(await new Par2Merger().mergeSet(setId), 0)
+    assert.equal((await db.from('releases').whereIn('id', ids).select('id')).length, 6)
+  })
+
   test('a set file that matches several releases nearby is ambiguous: not merged', async ({ assert }) => {
     const at = new Date('2022-07-08T03:46:00Z')
     const ids: number[] = []