1. Capture the symptom.
Note the message type, time range and affected recipient provider. Use a safe summary initially rather than forwarding private message bodies or headers.
Process
Follow one message before changing the whole system.
Identify the email, the expected destination and what the available records actually show.
2. Trace the sending path.
Review the application event, queue, provider response and later events. Identify missing records and recent changes.
3. Review the relevant configuration.
Check the sender, approved sending services and domain setup. Choose the smallest change supported by the evidence.
4. Test with permission.
Use isolated tests and approved destinations. Record provider acceptance separately from confirmed recipient arrival and inbox placement.
5. Hand over the result.
Document the changed path, failure handling, evidence and remaining limits. A cleared alert alone does not prove an earlier message was replayed or delivered.