Failure and Recovery Testing

Failure and Recovery Testing

Record how failure behaviour and recovery are verified.

Status: Partial implementation reference — Code-confirmed; operational validation pending

Last reviewed: 2026-10-19

Scope

Failure scenarios, fault injection where applicable, recovery checks and observed results. The source-confirmed baseline below covers part of this scope; remaining operational details are unknown.

Current implementation baseline

Existing source cases exercise rollback, retry identity, archive interruptions, lineage disagreement, queue/cursor partial failures and SQLite transaction recovery. The six Worker specifications retain Not run status for their documentation review; historical implementation reports separately record earlier runs.

Evidence: Corpus archive tests, lineage tests, Events pump tests and SDK sync tests. A local failure-injection test is not a cloud restore drill or proof against every crash schedule.

The backup restore procedure remains an acceptance drill with unresolved provider-backed work. Keep missing side-effect/concurrency coverage and the proposed cases in the inventory visible. No application test or recovery script was run during this documentation audit.