Implementation Notes for Recovery Checks That Happen Before Retry

Recovery checks should happen before retry because repeated action against broken state is often the shortest path to a larger incident.

Retry logic often starts too early. The safer order is to confirm the state, confirm the policy, and only then decide whether the action should happen again.

Recovery checks should verify

• State version integrity.
• Approval or policy validity.
• Whether partial side effects already escaped.

Use the quickstart, architecture guide, and examples to move recovery checks ahead of retries.