openclaw - 💡(How to fix) Fix Session file lock not released on abnormal exit, blocking subsequent requests

Bug Description

When an agent process terminates abnormally (e.g. crash, kill, network timeout), the session .lock file is not cleaned up. This causes subsequent requests to the same session to fail with:

Agent failed before reply: session file locked (timeout 60000ms): pid=XXXX /path/to/session.jsonl.lock

Steps to Reproduce

1. Agent process holding a session lock dies unexpectedly
2. Next message to the same session arrives
3. Lock file still exists, new process cannot acquire lock within 60s timeout
4. Request fails

Expected Behavior

• Stale lock files should be detected and handled automatically (e.g. lock expires after N seconds, or detect that the holding PID no longer exists)
• Alternatively, add a cleanup hook on process exit (SIGTERM/SIGINT) to release locks

Workaround

Manually delete the .lock file:

rm /path/to/session.jsonl.lock

Environment

• OpenClaw: latest (npm global install)
• OS: macOS 15.5 (arm64)
• Node: v26.0.0

Suggested Fix

1. PID-based stale detection: Before waiting for a lock, check if the PID in the lock file is still running. If not, treat the lock as stale and remove it.
2. Lock timeout: Add a max age to lock files (e.g. 5 minutes). If a lock is older than the threshold, consider it stale.
3. Process exit hooks: Register cleanup handlers for SIGTERM/SIGINT/SIGBREAK to release locks on graceful shutdown.