Kova OpenClaw Runtime Findings

Run: kova-2026-05-03T130727Z
Target: local-build:/home/runner/_work/openclaw/openclaw
Mode: execution
Platform: linux 6.5.13 (x64)

Failure Brief

Decision: FAIL
Primary blocker: agent-cold-warm-message/mock-openai-provider
Why: ocm @'kova-agent-cold-warm-message-mock-openai-provider-r1-kova-2026-05-03t130727z' -- agent --local --agent main --session-id kova-agent-cold-warm --message 'Reply with exact ASCII text KOVA_AGENT_OK only.' --thinking off --timeout 120 --json took 59496ms, over threshold 45000ms
Evidence:
- peakRssMb: 1153.6
- cpuPercentMax: 247.4
- coldAgentTurnMs: 59496
- warmAgentTurnMs: 36176
- agentColdWarmDeltaMs: 23320
- cold provider was fast (6ms), but OpenClaw spent 59295ms before provider work.
- agent-cli: 1153.6MB RSS, 247.4% CPU
- agent-process: 1153.6MB RSS, 247.4% CPU
Likely owner: OpenClaw
Paste to fixer:
Investigate OpenClaw release gate failure agent-cold-warm-message/mock-openai-provider. Kova decision was FAIL on linux-x64. Primary evidence: ocm @'kova-agent-cold-warm-message-mock-openai-provider-r1-kova-2026-05-03t130727z' -- agent --local --agent main --session-id kova-agent-cold-warm --message 'Reply with exact ASCII text KOVA_AGENT_OK only.' --thinking off --timeout 120 --json took 59496ms, over threshold 45000ms. Measurements: peakRssMb: 1153.6; cpuPercentMax: 247.4; coldAgentTurnMs: 59496; warmAgentTurnMs: 36176; agentColdWarmDeltaMs: 23320; cold provider was fast (6ms), but OpenClaw spent 59295ms before provider work.; agent-cli: 1153.6MB RSS, 247.4% CPU; agent-process: 1153.6MB RSS, 247.4% CPU. Dependency/plugin load errors were zero, so focus on startup, memory, CPU, gateway readiness, runtime deps staging, provider/model load, and UI asset initialization. Likely owner area: OpenClaw.

Recommended next scenario

Reason: ocm @'kova-agent-cold-warm-message-mock-openai-provider-r1-kova-2026-05-03t130727z' -- agent --local --agent main --session-id kova-agent-cold-warm --message 'Reply with exact ASCII text KOVA_AGENT_OK only.' --thinking off --timeout 120 --json took 59496ms, over threshold 45000ms
Command: node bin/kova.mjs run --target local-build:/home/runner/_work/openclaw/openclaw --scenario agent-cold-warm-message --state mock-openai-provider --execute --profile-on-failure --retain-on-failure --json

Scenario: agent-cold-warm-message
Result: FAIL
Cleanup: destroyed
Measurements:
- startup: listening unknown; health unknown; readiness unknown; gateway disabled; restarts 0
- resources: peak RSS 1153.6 MB; max CPU 247.4%; samples 110; roles agent-cli 1153.6MB/247.4%, agent-process 1153.6MB/247.4%, command-tree 1153.6MB/247.4%, package-manager 442.1MB/146%
- agent: turn 59496ms; cold/warm 59496ms/36176ms; cold-warm delta 23320ms; pre-provider 59295ms; provider 6ms; cleanup unknown; diagnosis pre-provider-stall; leaks 0
- plugins/runtime: missing deps 0; plugin failures 0; runtime deps unknown; warm restages unknown; warm reuse unknown
- diagnostics: timeline unavailable; slowest span unknown unknown; embedded traces 0; liveness warnings 0; open spans 0 (0 required); node CPU/heap/trace 0/0/0
Violations:
- ocm @'kova-agent-cold-warm-message-mock-openai-provider-r1-kova-2026-05-03t130727z' -- agent --local --agent main --session-id kova-agent-cold-warm --message 'Reply with exact ASCII text KOVA_AGENT_OK only.' --thinking off --timeout 120 --json took 59496ms, over threshold 45000ms
- peak RSS 1153.6 MB exceeded threshold 900 MB
- agent-cli peak RSS 1153.6 MB exceeded threshold 900 MB
- agent-process peak RSS 1153.6 MB exceeded threshold 900 MB
- OpenClaw diagnostics timeline was required for profile 'diagnostic' on target kind 'local-build' but was not emitted
- cold agent spent 59295ms before provider work, over threshold 10000ms
- warm agent spent 36043ms before provider work, over threshold 10000ms
- cold agent turn took 59496ms, over threshold 45000ms
- warm agent turn took 36176ms, over threshold 15000ms
- cold pre-provider latency was 59295ms, over threshold 10000ms
- warm pre-provider latency was 36043ms, over threshold 10000ms
- cold provider was fast (6ms), but OpenClaw spent 59295ms before provider work.

Scenario: agent-cold-warm-message
Result: FAIL
Cleanup: destroyed
Measurements:
- startup: listening unknown; health unknown; readiness unknown; gateway disabled; restarts 0
- resources: peak RSS 1044.6 MB; max CPU 248.4%; samples 110; roles agent-cli 1044.6MB/248.4%, agent-process 1044.6MB/248.4%, command-tree 1044.6MB/248.4%, package-manager 419.3MB/147%
- agent: turn 58733ms; cold/warm 58733ms/37167ms; cold-warm delta 21566ms; pre-provider 58593ms; provider 1ms; cleanup unknown; diagnosis pre-provider-stall; leaks 0
- plugins/runtime: missing deps 0; plugin failures 0; runtime deps unknown; warm restages unknown; warm reuse unknown
- diagnostics: timeline unavailable; slowest span unknown unknown; embedded traces 0; liveness warnings 0; open spans 0 (0 required); node CPU/heap/trace 0/0/0
Violations:
- ocm @'kova-agent-cold-warm-message-mock-openai-provider-r2-kova-2026-05-03t130727z' -- agent --local --agent main --session-id kova-agent-cold-warm --message 'Reply with exact ASCII text KOVA_AGENT_OK only.' --thinking off --timeout 120 --json took 58733ms, over threshold 45000ms
- peak RSS 1044.6 MB exceeded threshold 900 MB
- agent-cli peak RSS 1044.6 MB exceeded threshold 900 MB
- agent-process peak RSS 1044.6 MB exceeded threshold 900 MB
- OpenClaw diagnostics timeline was required for profile 'diagnostic' on target kind 'local-build' but was not emitted
- cold agent spent 58593ms before provider work, over threshold 10000ms
- warm agent spent 37029ms before provider work, over threshold 10000ms
- cold agent turn took 58733ms, over threshold 45000ms
- warm agent turn took 37167ms, over threshold 15000ms
- cold pre-provider latency was 58593ms, over threshold 10000ms
- warm pre-provider latency was 37029ms, over threshold 10000ms
- cold provider was fast (1ms), but OpenClaw spent 58593ms before provider work.

Scenario: agent-cold-warm-message
Result: FAIL
Cleanup: destroyed
Measurements:
- startup: listening unknown; health unknown; readiness unknown; gateway disabled; restarts 0
- resources: peak RSS 1154.9 MB; max CPU 234.5%; samples 105; roles agent-cli 1154.9MB/234.5%, agent-process 1154.9MB/234.5%, command-tree 1154.9MB/234.5%, package-manager 438.2MB/145%
- agent: turn 57007ms; cold/warm 57007ms/34150ms; cold-warm delta 22857ms; pre-provider 56890ms; provider 2ms; cleanup unknown; diagnosis pre-provider-stall; leaks 0
- plugins/runtime: missing deps 0; plugin failures 0; runtime deps unknown; warm restages unknown; warm reuse unknown
- diagnostics: timeline unavailable; slowest span unknown unknown; embedded traces 0; liveness warnings 0; open spans 0 (0 required); node CPU/heap/trace 0/0/0
Violations:
- ocm @'kova-agent-cold-warm-message-mock-openai-provider-r3-kova-2026-05-03t130727z' -- agent --local --agent main --session-id kova-agent-cold-warm --message 'Reply with exact ASCII text KOVA_AGENT_OK only.' --thinking off --timeout 120 --json took 57007ms, over threshold 45000ms
- peak RSS 1154.9 MB exceeded threshold 900 MB
- agent-cli peak RSS 1154.9 MB exceeded threshold 900 MB
- agent-process peak RSS 1154.9 MB exceeded threshold 900 MB
- OpenClaw diagnostics timeline was required for profile 'diagnostic' on target kind 'local-build' but was not emitted
- cold agent spent 56890ms before provider work, over threshold 10000ms
- warm agent spent 34012ms before provider work, over threshold 10000ms
- cold agent turn took 57007ms, over threshold 45000ms
- warm agent turn took 34150ms, over threshold 15000ms
- cold pre-provider latency was 56890ms, over threshold 10000ms
- warm pre-provider latency was 34012ms, over threshold 10000ms
- cold provider was fast (2ms), but OpenClaw spent 56890ms before provider work.
