Duplicate and Drop Audit (QA)

Duplicate and Drop Audit (QA)

1. 01

   Pull the last 30 days of delivery logs and count total inbound events against total downstream deliveries

   A gap between the two counts is the first signal of silent drops. Log the delta as a percentage before touching any configuration.

2. 02

   Replay a sample of 50 events with known unique keys and confirm each one lands exactly once

   EventSend deduplicates on optional unique keys, so a missing key on the client side produces duplicates that look like platform failures. Check the payload, not just the endpoint.

3. 03

   Trace every retry that fired more than twice and record the final HTTP status

   Repeated retries against a 4xx response usually mean a malformed payload or a revoked destination token. Retries against 5xx are the destination's problem, not the routing layer's.

4. 04

   Verify each destination in the fan-out list still resolves and accepts the current payload shape

   Slack, Discord, Telegram, and custom webhook endpoints change auth schemes without notice. A destination that silently rejects is worse than one that errors loudly.

5. 05

   Confirm the client's event naming convention has not drifted since the last audit

   Renamed or newly added event types that were never mapped to a destination will pass through the endpoint and vanish. Diff the current event list against the original mapping document.

6. 06

   Document every unresolved duplicate or drop with the event type, timestamp, and affected destination

   This log becomes the evidence base for the next client review and the input for any retry or reconciliation work you bill for.

7. 07

   Set a recurring 30-day re-audit date and assign an owner inside the delivery team

   Webhook reliability degrades quietly as client stacks change. A standing calendar entry is the only thing that catches drift before the client does.