Summary

kanban_db.py initializes SQLite with WAL journal mode and synchronous=NORMAL, but does NOT set busy_timeout. When two processes (Gateway Dispatcher + Dashboard Plugin) attempt simultaneous writes, SQLite returns SQLITE_BUSY instead of waiting. This causes WAL journal corruption.

One-line fix verified stable over 12+ hours:

conn.execute("PRAGMA busy_timeout=10000")

Environment

• Hermes Agent v0.14.0 (2026.5.16) on Ubuntu 24.04
• 6 active Hermes profiles, Kanban with 30+ tasks
• Gateway dispatcher + Dashboard Kanban plugin concurrently

Steps to Reproduce

1. hermes kanban with dispatch_in_gateway: true, 30+ tasks
2. hermes dashboard --tui concurrently
3. Wait 10-30 min
4. hermes kanban list → database disk image is malformed

Root Cause

kanban_db.py line 1182-1185 sets WAL + synchronous but no busy_timeout. Concurrent writes get immediate SQLITE_BUSY, abort mid-transaction, corrupting WAL file.

Related

• kanban.db corruption recurs under concurrent reclaim-SIGKILL even with synchronous=FULL + wal_autocheckpoint=100 #31618 (same corruption root)
• fix(kanban): close leaked decompose connections + don't mask I/O error on rollback #32415 (leaked decompose connections compound)
• fix: harden kanban corrupt board quarantine #32094 (quarantine hardening)