openclaw - 💡(How to fix) Fix [Bug]: Gateway CPU pegged at 100%+ continuously; gateway.err.log grows to 18-19 GB

Bug type

Behavior bug (incorrect output/state without crash)

Beta release blocker

No

Summary

Gateway process pegs CPU at 100–107% continuously due to CopyDataPropertiesWithExcludedProperties in a hot async microtask loop (object spread); gateway.err.log grows to 18–19 GB from a dep-staging error loop.

Steps to reproduce

1. Start OpenClaw gateway normally.
2. Observe CPU usage — gateway process climbs to 100–107% and stays there indefinitely.
3. Check ~/.openclaw/logs/gateway.err.log — file grows continuously (observed: 18–19 GB) from a repeating error loop involving acpx/runway/tts-local-cli dep-staging.
4. Inter-agent communication (e.g., main ↔ alchemy subagent) becomes sluggish or fails.
5. Workaround: openclaw gateway restart — CPU drops immediately. Bug recurs without patching.

Expected behavior

Gateway CPU usage stays at normal idle levels (<5%). Log file grows slowly with normal operational output only.

Actual behavior

Gateway process (node) continuously pegged at 100–107% CPU as observed in Activity Monitor. Stack trace shows CopyDataPropertiesWithExcludedProperties in async microtask queue — consistent with object spread ({...obj}) inside a hot async loop. Separately, gateway.err.log grows unbounded (18–19 GB observed across two incidents) from a tight error loop involving acpx/runway/tts-local-cli dep-staging. Both symptoms resolve immediately on gateway restart but recur on next launch.

OpenClaw version

2026.5.7 (eeef486)

Operating system

macOS 25.3.0 (Darwin arm64) — Apple Silicon Mac mini

Install method

npm global (homebrew node v22.22.0)

Model

anthropic/claude-sonnet-4-6

Provider / routing chain

openclaw gateway -> anthropic (direct)

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

Affected: Any multi-agent OpenClaw setup with gateway running long-term Severity: High — pegged CPU degrades system performance and causes inter-agent communication failures Frequency: Always (observed twice, reproduces on every gateway launch) Consequence: System becomes sluggish, inter-agent sessions fail or timeout, disk can fill from log growth (18–19 GB observed); requires manual gateway restart to recover

Additional information

Two separate incidents observed (2026-05-02 and 2026-05-08). Both recovered immediately on openclaw gateway restart. Bug recurs on next launch without a code fix.

Two distinct symptoms that may have separate root causes:

1. CPU loop: CopyDataPropertiesWithExcludedProperties visible in node profiler / Activity Monitor call stack — likely object spread in a hot async loop within the gateway.
2. Log bloat: Tight error loop from acpx/runway/tts-local-cli dep-staging writing continuously to gateway.err.log — no log rotation or backpressure applied.

Workaround: openclaw gateway restart clears both symptoms until next recurrence.