fix: best-effort HNSW thread-pin retrofit + drop dead attempt-cap constant

Addresses remaining PR #976 review items after rebase on develop.

`get_collection(create=False)` previously returned existing collections without
re-applying `hnsw:num_threads=1`, so palaces created before the fix kept the
unsafe parallel-insert path. Add `_pin_hnsw_threads()` helper that calls
`collection.modify(configuration=UpdateCollectionConfiguration(
hnsw=UpdateHNSWConfiguration(num_threads=1)))` best-effort on every
`get_collection` call (including the MCP server's `_get_collection`).

In chromadb 1.5.x the runtime config does not persist to disk across
`PersistentClient` reopens, so the retrofit is re-applied each process start
rather than being a one-shot migration. Fresh palaces keep the metadata-based
pin as primary defense; legacy palaces now also get per-session protection
without requiring `mempalace nuke` + re-mine.

After the rebase on develop, `hook_precompact` delegates to `_mine_sync` and
no longer emits `decision: block`, so the attempt-cap constant was orphaned.
Grep confirms 0 usages in the repo — remove it.

- `_pin_hnsw_threads` retrofits legacy collection (num_threads None -> 1)
- `_pin_hnsw_threads` swallows all errors (never raises)
- `ChromaBackend.get_collection(create=False)` applies retrofit on legacy palace
- 62 tests pass (10 backends + 6 palace locks + 46 hooks_cli)

mempalace/backends/chroma.py

@@ -130,6 +130,37 @@ def quarantine_stale_hnsw(palace_path: str, stale_seconds: float = 3600.0) -> li
     return moved
 
 
+def _pin_hnsw_threads(collection) -> None:
+    """Best-effort retrofit: pin ``hnsw:num_threads=1`` on an existing collection.
+
+    Fresh collections set this via ``metadata=`` at creation. Legacy palaces
+    built before that change keep the default (parallel insert) and can hit
+    the HNSW race described in #974/#965. ChromaDB's
+    ``collection.modify(configuration=...)`` lets us re-apply ``num_threads=1``
+    in memory at load time so every new process is protected.
+
+    Note: in chromadb 1.5.x the modified ``configuration_json["hnsw"]`` does
+    not persist to disk across ``PersistentClient`` reopens, so this must
+    run on every ``get_collection`` call, not just once.
+    """
+    try:
+        from chromadb.api.collection_configuration import (
+            UpdateCollectionConfiguration,
+            UpdateHNSWConfiguration,
+        )
+    except ImportError:
+        logger.debug("_pin_hnsw_threads skipped: chromadb too old", exc_info=True)
+        return
+    try:
+        collection.modify(
+            configuration=UpdateCollectionConfiguration(
+                hnsw=UpdateHNSWConfiguration(num_threads=1)
+            )
+        )
+    except Exception:
+        logger.debug("_pin_hnsw_threads modify failed", exc_info=True)
+
+
 def _fix_blob_seq_ids(palace_path: str) -> None:
     """Fix ChromaDB 0.6.x -> 1.5.x migration bug: BLOB seq_ids -> INTEGER.
 
@@ -572,6 +603,7 @@ class ChromaBackend(BaseBackend):
             )
         else:
             collection = client.get_collection(collection_name, **ef_kwargs)
+        _pin_hnsw_threads(collection)
         return ChromaCollection(collection)
 
     def close_palace(self, palace) -> None:

mempalace/hooks_cli.py

@@ -643,9 +643,6 @@ def hook_session_start(data: dict, harness: str):
     _output({})
 
 
-MAX_PRECOMPACT_BLOCK_ATTEMPTS = 2
-
-
 def hook_precompact(data: dict, harness: str):
     """Precompact hook: mine transcript synchronously, then allow compaction."""
     parsed = _parse_harness_input(data, harness)

mempalace/mcp_server.py

@@ -57,7 +57,7 @@ from .config import (  # noqa: E402
     sanitize_content,
 )
 from .version import __version__  # noqa: E402
-from .backends.chroma import ChromaBackend, ChromaCollection  # noqa: E402
+from .backends.chroma import ChromaBackend, ChromaCollection, _pin_hnsw_threads  # noqa: E402
 from .query_sanitizer import sanitize_query  # noqa: E402
 from .searcher import search_memories  # noqa: E402
 from .palace_graph import (  # noqa: E402
@@ -219,20 +219,23 @@ def _get_collection(create=False):
         if create:
             # hnsw:num_threads=1 disables ChromaDB's multi-threaded ParallelFor
             # HNSW insert path, which has a race in repairConnectionsForUpdate /
-            # addPoint (see issues #974, #965). The setting is only honored at
-            # collection creation time — pre-existing palaces created before
-            # this fix keep the unsafe default; users must `mempalace nuke` +
-            # re-mine to get the protection on legacy palaces.
-            _collection_cache = ChromaCollection(
-                client.get_or_create_collection(
-                    _config.collection_name,
-                    metadata={"hnsw:space": "cosine", "hnsw:num_threads": 1},
-                )
+            # addPoint (see issues #974, #965). Set via metadata on fresh
+            # collections and re-applied via _pin_hnsw_threads() for legacy
+            # palaces whose collections were created before this fix (the
+            # runtime config does not persist cross-process in chromadb 1.5.x,
+            # so the retrofit runs every time _get_collection opens a cache).
+            raw = client.get_or_create_collection(
+                _config.collection_name,
+                metadata={"hnsw:space": "cosine", "hnsw:num_threads": 1},
             )
+            _pin_hnsw_threads(raw)
+            _collection_cache = ChromaCollection(raw)
             _metadata_cache = None
             _metadata_cache_time = 0
         elif _collection_cache is None:
-            _collection_cache = ChromaCollection(client.get_collection(_config.collection_name))
+            raw = client.get_collection(_config.collection_name)
+            _pin_hnsw_threads(raw)
+            _collection_cache = ChromaCollection(raw)
             _metadata_cache = None
             _metadata_cache_time = 0
         return _collection_cache