Bug

After a SIGUSR1-based gateway restart (triggered by config.patch, gateway restart, or openclaw update), the launchd service frequently becomes unloaded. Subsequent openclaw gateway restart or openclaw gateway stop && openclaw gateway start commands fail with "service not loaded."

Root Cause (suspected)

The gateway restart sends SIGUSR1, causing the process to exit and re-launch. When multiple restarts happen in quick succession (e.g. several config.patch calls), launchd interprets the rapid exits as a crash loop and silently unloads the service via its built-in throttling. Once unloaded, no restart commands work because launchd no longer knows about the service.

Reproduction

1. Run several config.patch calls in a row (or multiple openclaw gateway restart)
2. After 2-3 restarts, run openclaw gateway restart → fails with "service not loaded"
3. launchctl list | grep openclaw shows no entry

Workaround

openclaw gateway install && openclaw gateway restart

gateway install rewrites the plist and re-registers it with launchctl load/launchctl bootstrap, which always fixes it.

Expected Behavior

• SIGUSR1 restarts should not cause launchd to unload the service
• openclaw update should automatically re-register the launchd service (run gateway install internally)
• Ideally, the restart mechanism should handle launchd's crash-loop throttling (e.g. add a small delay, or use a different restart approach)

Environment

• OS: macOS 15 (Sequoia), Apple Silicon
• OpenClaw: 2026.2.9
• Node: v25.5.0
• Plist: ~/Library/LaunchAgents/ai.openclaw.gateway.plist with KeepAlive: true