Trigger
Cloud scheduler
A platform schedule starts the bounded cycle every minute.
* * * * *Public continuity evidence
Runs without an open browser.
The bounded automation slice runs on Cloudflare Cron and a transactional D1 queue with leases, retries and version-bound heartbeats. It does not depend on a browser tab, a local process or the manual run route.
60seconds between scheduled ticks
—version-bound live observations
—queue items due at verification
0browser, local process or manual route dependencies
Truth boundary
Browser independence is verified for this slice. Full continuity is not.
The scheduler is currently degraded because provider recovery is still pending. This page does not prove 24-hour continuity, provider recovery, queue stress, disaster recovery, every product workflow or a score of 99.
01 / Durable runtime
Cloud scheduling and durable queue state replace an open tab.
The proof is tied to one serving Worker, one rollback Worker and immutable database observations.
State
Transactional queue
Queue entries, locks, lease expiry and retries are stored in D1.
Health
Degraded, not hidden
The evidence preserves the current provider failure signal instead of presenting the runtime as healthy.
02 / Live observations
Three advancing heartbeats from the exact serving version.
No exact-version continuity sequence is available for this response.
03 / Open gates
What remains before a broader continuity claim.
- Gate 1Observe the exact production Worker for a continuous 24-hour window.
- Gate 2Restore the degraded discovery provider lane and verify recovery without hiding the prior failures.
- Gate 3Run bounded queue saturation, lease recovery and replay tests against a production-shaped isolated environment.
- Gate 4Complete a disaster-recovery exercise for the scheduler and queue state.
- Gate 5Audit every product workflow before extending this bounded continuity claim to the whole product.