openclaw - 💡(How to fix) Fix Feature request: configurable RSS memory pressure threshold

Summary

The gateway's memory pressure monitor has a hardcoded RSS threshold of 1,610,612,736 bytes (1.5GB). There is no config key to change this value.

Problem

For users running OpenClaw on high-memory machines with NODE_OPTIONS=--max-old-space-size=12288 (12GB V8 heap ceiling), the 1.5GB RSS threshold fires constantly even during normal fleet operation — especially when multiple isolated cron jobs spawn concurrently. This causes:

• Frequent [diagnostics/memory] memory pressure: level=warning reason=rss_threshold log spam
• Isolated cron sessions being aborted mid-run via EmbeddedAttemptSessionTakeoverError when under memory pressure
• The 1.5GB threshold bears no relationship to the configured heap ceiling or actual available system RAM

Expected behavior

A config key such as diagnostics.memoryPressureRssThresholdBytes (or similar) that allows operators to set the RSS warning threshold appropriate for their machine's RAM.

Suggested schema

{
  "diagnostics": {
    "memoryPressureRssThresholdBytes": 4294967296
  }
}

Or alternatively, derive the threshold as a fraction of the configured --max-old-space-size value (e.g. default to 50% of V8 heap ceiling rather than a hardcoded 1.5GB).

Environment

• OpenClaw version: 2026.5.27
• Host: macOS, Apple Silicon, 64GB RAM
• NODE_OPTIONS=--max-old-space-size=12288 set via service env
• Fleet running 40+ agents with multiple concurrent isolated cron jobs