openclaw-test-heap-leaks

Compare V8 heap snapshots across time within the same Vitest worker PID to identify retained object families driving memory growth. Use the provided delta tool to surface the largest snapshot deltas by self size while applying thresholds to reduce noise and isolate genuine leaks.

Test-induced heap growth in parallel test lanes is typically caused by either shared-worker retained transformed modules or real runtime cleanup and lifecycle leaks. Comparing V8 heap snapshots from the same worker PID helps distinguish between scheduling hotspot retention and genuine cleanup logic failures.

Run tests with heap snapshot intervals enabled, then use heapsnapshot-delta.mjs to compare snapshots from the specific lane directory and worker PID that reproduced the growth. Apply retention and dominator confirmation to validate the identified object families driving the leak.

Yes, lane-aware isolation is required because it preserves the real test lane and shape, such as worker, shard, or batch lane directories. This ensures snapshot comparisons reflect the same execution context and prevents cross-lane data contamination.

Compare V8 heap snapshots across time within the same worker PID to classify the leak source. Route investigation toward scheduling or hotspot isolation for transformed-module retention, or toward cleanup logic and lifecycle fixes for genuine runtime leaks.