The two-order AWS experiment
One product.
Two orders. Real retries.
Three choices of key face the same failure: a receipt commits, the fulfillment response fails, and standard SQS delivers the message again.
No verified recording has been loaded.
A successful experiment can expose violating handlers. Their failures stay visible.
Deletion has not been established by this page.
Independent DynamoDB observations
Every valid order needs a receipt.
The count comes from complete recorded ledger queries, independently of the worker’s return value.
| Valid order | No keyEvery attempt | Product keysku | Order keyorderId |
|---|---|---|---|
| Receipt snapshot | —Not observed | —Not observed | —Not observed |
| Business result | ◌ Incomplete | ◌ Incomplete | ◌ Incomplete |
A receipt is the entire simulated fulfillment. These observations do not establish safety for a real payment or shipment, every SQS schedule, or all future deliveries.
A controlled starting point
Observe A.
Then submit B.
Three original messages submit A, one per handler. The runner waits until A’s receipt and correlated post-commit failure are observed for every handler before submitting B.
SQS decides the actual retry schedule. The page shows that recorded schedule without forcing it into the local replay’s four steps.
- 01A’s receipt commits
The receiver owns the durable receipt and conditionally creates it.
- 02The response fails after commit
The first receive fails before returning success. The receipt survives.
- 03A’s evidence opens the B gate
B is a different valid order sharing the same product.
- 04Observe real redeliveries
Keep all observed receipts, conflicts and retries. Missing evidence stays incomplete.
Captured worker and receiver events
Follow one handler’s evidence.
No recorded events are available yet.
- Waiting for a real retained AWS report.
Event timestamps describe the supplied recording. Cross-service timestamps do not establish a total execution order.
Checks you can inspect and run
A result with its reasons.
The full Python verifier is rerun against the report during the site build. This browser checks the report’s SHA-256 and recomputes counts from the ledger reads. Neither proves who generated the export.
Inspect the verifier checks No result
No report has been verified.