Skip to content

orb(disposition): maybeRunAgentMaintenance silently drops the disposition on lock contention — no retry, no audit #9025

Description

@JSONbored

src/queue/processors.ts ~2447-2450: if (!actuationLock.acquired) return; — the pass returns, the job completes "successfully", and nothing re-queues the disposition. No audit event records that a planned action was abandoned.

Contrast review-evasion.ts ~80-82, which throws the retryable PrActuationLockContendedError (5s backoff) for the same condition. The class exists precisely for this.

This is a silent amplifier for every restart incident: after a kill, the recovered job re-runs, republishes a no-op surface, then hits its own dead predecessor's orphaned actuation lock here and returns — so the disposition is lost a second time, now with no trace at all.

Fix

Throw PrActuationLockContendedError here too (or at minimum audit the abandonment and enqueue a trailing agent-regate-pr).

Refs #9003, #9007, #9021.

Metadata

Metadata

Assignees

Labels

maintainer-onlyOwner-only work — yields no Gittensor points.orbGittensory Orb related - maintainer self-hosting analytics.

Projects

No projects

Relationships

None yet

Development

No branches or pull requests

Issue actions