ClaraWideIncident notification delivery QA
v0.12.27 · DELIVERY RELIABILITY FOUNDATION

A security notification should either arrive—or fail loudly with evidence.

This release adds durable delivery jobs, exclusive worker leases, bounded retry/backoff, expired-lease recovery, dead-letter handling and immutable attempt evidence. A real email/pager/webhook provider and production schedule remain intentionally gated; v0.12.28 adds the separate worker scheduling/backpressure control layer.

SCHEMA + ATTACK PATHS

Automated delivery checks

Confirms schema v26, one-job-per-notification enqueue, exclusive leases, retry success, retry exhaustion, dead-letter evidence, stale-lease recovery and completion-bypass blocking.

NOT RUNSchema check has not run yet.
NOT RUNSelf-test has not run yet.
VISIBLE RETRY FLOW

Queue → lease → retry → deliver / dead-letter

Use the staging simulator to watch the durable job move through the same states a future production delivery worker will use.

READYStart with a QA Delivery Admin session.
DURABLE DELIVERY JOBS

Worker queue state

Queued/retry work stays visible until it is delivered, suppressed, or exhausts its bounded retry budget.

No QA delivery dashboard loaded yet.

ATTEMPT EVIDENCE

Immutable delivery attempts

No attempt evidence loaded yet.

DEAD-LETTER QUEUE

Exhausted notifications

No dead-letter records loaded yet.

PRODUCTION GATE — DO NOT FAKE

Production delivery evidence

These stay open until real scheduling, provider credentials, delivery callbacks, dead-letter alerting and operations monitoring actually exist.

Recent delivery QA runs

No QA runs recorded yet.

Evidence tools

Clearing browser evidence does not rewrite server-owned delivery jobs, attempt records, worker runs or dead-letter records.