A webhook can arrive more than once. A worker can also crash after changing an external system but before returning a response. Webhook Retry Lab lets you compare the consequences before you choose a recovery pattern.
Free, no account required. Runs in your browser without sending real webhooks, API keys or customer payloads.
Try three failure patterns
- Duplicate action: replay one event without deduplication and watch extra business actions accumulate.
- Lost event: write a seen flag, then crash before the action. A retry can skip work that never happened.
- Recovered event: compare a state ledger with a recoverable lease and a destination that accepts a stable idempotency key.
Change the crash boundary and delivery count, compare all three patterns, then download a JSON scenario report. The report describes the educational model; it is not a production audit.
Use the result in your n8n design
After trying the lab, follow our guides to build an idempotency gate and return quickly with a durable async job pattern.
Limits and optional statistics
The lab assumes sequential deliveries, one fixed event, a recoverable lease that expires before the next retry, and destination idempotency that persists for the scenario. It does not test concurrency, real databases, key expiry or your production configuration. One action in the model does not prove exactly-once delivery.
Optional anonymous statistics count visits, runs, completed simulations, download requests, errors and guide clicks by broad traffic source. You can keep statistics off and still use every feature. Names, emails, IP addresses, identifiers and scenario settings are not stored in the application statistics database. The hosting provider processes connection data to deliver the service.