A trigger event can be delivered again, and an operator can retry a failed run. A flow that always creates a new row can therefore duplicate the same business request. Use a stable business key and inspect the destination before deciding whether to create or update. The destination's uniqueness controls matter as well as the flow's checks.
A check-then-create pattern is useful for sequential local practice, but simultaneous runs can both observe that a key is absent. A workplace design needs the data store's uniqueness or the API's idempotency mechanism when concurrency is possible. This lesson checks the visible effects for a new request and for a repeated request. A timeout after a write creates uncertainty rather than proof that nothing happened. Reconcile the key and destination before repeating a side effect.