ROOT CAUSE: Monitor was doing DROP DATABASE when NODE2 agents were missing,
but the backup didn't have NODE2 agents, causing an infinite loop.
FIX:
- FULL RECOVERY (DROP DATABASE) only when MicroDAOs < 5 (critical data loss)
- SOFT RECOVERY (just sync agents) when MicroDAOs exist but agents missing
- Prefer backup with NODE2 agents (full_backup_with_node2*.sql)
- Never DROP DATABASE if MicroDAOs exist
This prevents the daily data loss issue.
- Check for at least 45 NODE2 agents (out of 50 expected)
- This prevents false positives when only core agents exist
- Better detection of actual data loss
- Add monitor-db-stability.sh for automatic recovery
- Improve PostgreSQL shutdown settings to prevent data loss
- Add checkpoint and WAL settings for better persistence
