Summary

Two related issues that together cause eventLoopUtilization=1 and cpuCoreRatio=1.5+, making the gateway unresponsive (cron tool timeouts, slow responses).

Issue 1: Temp files not cleaned up after atomic writes

.tmp files accumulate indefinitely after atomic writes:

• ~/.openclaw/devices/paired.json.*.tmp
• ~/.openclaw/cron/jobs-state.json.*.tmp
• ~/.openclaw/update-check.json.*.tmp

After ~2 weeks of runtime: 150+ paired.json temps, 300+ cron state temps, 67+ update-check temps.
These should be deleted after the atomic rename succeeds. The accumulation causes significant I/O overhead.

Issue 2: node.list probe takes ~2 seconds per call, Control UI polls continuously

With no nodes paired, node.list still performs a 2-second probe timeout on every call.
The Control UI dashboard polls this endpoint continuously via WebSocket (~every 2s).

Result: event loop is 100% utilized processing 2-second timeouts in a tight loop.

Log evidence:

node.list 2059ms (repeated every ~2s)
liveness warning: eventLoopUtilization=1 cpuCoreRatio=1.546

Suggested fixes

1. Temp files: delete .tmp after successful rename; add startup cleanup for orphaned temps
2. node.list: return immediately when no nodes configured, or make probe timeout configurable

Environment

• macOS arm64, Mac mini 8GB RAM
• OpenClaw 2026.4.29
• No nodes paired, iMessage + WhatsApp active