Operations, usage & worker pools

Follow a failed run

1. Open Operations for the correct Project.
2. Find the run and inspect node status and persisted attempts.
3. Use trace references and safe failure reasons to locate the boundary that failed.
4. Check approval state, worker availability and applicable quotas before resubmitting.
5. Use recovery or cancellation controls only when authorized and appropriate.

Workers and capacity

Assigned logical pools require eligible, healthy, non-draining worker registrations. A pool configuration is not proof that a worker is available. Waiting for a worker is different from provider failure. Leases distinguish execution attempts from later recovery.

Usage and auditing

Provider-reported usage, operational events, governance audit records and logs serve different purposes. Missing usage must remain unavailable rather than fabricated. Current metering is not billing-authoritative reconciliation, and application audit records are not independently tamper-proof.

Isolation boundary

Logical pool routing does not prove physical isolation or automatic scaling. Dedicated infrastructure, authenticated workload identity, network policy and production monitoring require deployment-specific implementation and verification.