Critical Data write loop, HNSW compaction bug

[raizenica]

┌───────────┬──────────────┬─────────────┐
│ Metric │ Before │ After │
├───────────┼──────────────┼─────────────┤
│ Used │ 1.7 T (100%) │ 211 G (13%) │
├───────────┼──────────────┼─────────────┤
│ Available │ 13 G │ 1.5 T │
└───────────┴──────────────┴─────────────┘

Root cause: my mempalace mine on the auto-memory dir hit a chromadb 1.x HNSW compaction bug that wrote a 10.2 TB sparse link_lists.bin (≈1.5 TB on disk) before crashing. The
crash didn't release the file handle, and the MCP server processes pinned the deleted inodes. Kill of PIDs 1581835 / 1581850 freed everything.

Environment:

• OS: Nobara
• MemPal version: 3.1.0