# Recover an uncertain write without creating a duplicate

1. Assign an operation key before sending the write and retain the exact payload.
2. When the response is lost, treat the outcome as unknown. Do not create a new operation automatically.
3. Retry the identical payload with the retained key when the endpoint supports it.
4. Read the returned resource ID through its normal detail endpoint.
5. Confirm its intended fields and attribution.
6. If the key conflicts with a different payload, stop and inspect the intended change.
7. Present success only after the committed result is known.