docs(README): Row 6 — cross-ref MemPalace#357 and MemPalace#1071 on MemPalace#1088, revise PR ordering

Author: jphein

README.md

@@ -29,7 +29,7 @@ Size (lines of diff) and Risk (maintainer-appetite + chance of a rework request)
 | **Reliability** | `quarantine_stale_hnsw()` should also validate the HNSW `index_metadata` file integrity — a process killed mid-write leaves a corrupt file with zero mtime drift, so the 1h threshold doesn't fire and the next palace open segfaults. Fix: attempt to deserialize the file in the quarantine check and quarantine on failure regardless of mtime. | Blocked on [#1062](https://github.com/MemPalace/mempalace/pull/1062) — @Samaara-Das's PR wires the quarantine caller that v3.3.2 shipped without (fixes [#1061](https://github.com/MemPalace/mempalace/issues/1061)). File our integrity-check extension as a follow-up after #1062 merges. | small | low | `backends/chroma.py` |
 | **CLI** | `cmd_purge` CLI — delete drawers by wing/room, nuke-and-reinsert retained drawers to avoid HNSW ghost entries left by `collection.delete()`. Substantive new command. | PR pending (file directly; reference [#521](https://github.com/MemPalace/mempalace/issues/521) for the hnswlib race that motivates the nuke-and-reinsert pattern) | medium | medium | `cli.py` |
 | **Hooks** | `hooks/mempal_precompact_hook.sh` transcript auto-mining — session ID parsed from JSON input, transcript resolved by direct path or `find`-by-session-id fallback, then mined inline (chunk_exchanges → upsert) before compaction fires; includes Python auto-detection | **On hold** — [#1083](https://github.com/MemPalace/mempalace/issues/1083) (@raphaelsamy, 2026-04-21) reports hook auto-mine with default flags produces polluted mega-wings; demanding opt-out. Our proposal adds *more* hook auto-mining in the same direction — possibly opposite to where maintainer wants to go. Engage #1083 before filing. | medium | low-medium | `hooks/mempal_precompact_hook.sh` |
-| **Performance** | `bulk_check_mined()` paginated pre-fetch + `--workers` ThreadPoolExecutor concurrent mining | [Issue #1088](https://github.com/MemPalace/mempalace/issues/1088) filed 2026-04-21 — shared benchmark data (~14 min → ~3 min on 165K-drawer palace), flagged intent overlap with merged [#784](https://github.com/MemPalace/mempalace/pull/784), asked if in-process fan-out is the right direction. PR to follow if maintainer approves. | medium | medium | `palace.py`, `miner.py` |
+| **Performance** | `bulk_check_mined()` paginated pre-fetch + `--workers` ThreadPoolExecutor concurrent mining | [Issue #1088](https://github.com/MemPalace/mempalace/issues/1088) filed 2026-04-21; [cross-ref comment](https://github.com/MemPalace/mempalace/issues/1088#issuecomment-4292570126) ties it to [#357](https://github.com/MemPalace/mempalace/issues/357) (parallel-mining corruption we could fix) and gates the PR on [#1071](https://github.com/MemPalace/mempalace/pull/1071) landing first (ORT thread cap, for bounded parallelism). | medium | medium | `palace.py`, `miner.py` |
 | **Reliability** | `_get_client()` tries `get_collection` before `create_collection` — `get_or_create_collection` segfaults ChromaDB 1.5.x when the existing collection's metadata differs from the call-site metadata | PR candidate (issue-first — sibling to [#974](https://github.com/MemPalace/mempalace/issues/974) @vince-mod3's `num_threads` SIGSEGV; frame ours so maintainer can sequence both chromadb-segfault workarounds together) | small | medium | `backends/chroma.py` |
 | **Reliability** | Skip `_fix_blob_seq_ids` sqlite open after first successful migration via `.blob_seq_ids_migrated` marker — opening sqlite3 against a live ChromaDB 1.5.x file corrupts the next PersistentClient | PR candidate (issue-first: narrow chromadb 1.5.x debugging path) | small | medium | `backends/chroma.py` |
 | **Performance** | Graph cache — 60s TTL, invalidated on writes | [#661](https://github.com/milla-jovovich/mempalace/pull/661) | small | medium | `palace_graph.py` |