A retry can help a transient failure, but an unbounded loop can repeat work and exceed the task budget. Distinguish a retryable response from a permanent request error. Respect a supplied retry delay and use backoff with jitter when the policy permits another attempt. In this lesson you configure a bounded retry policy for a mock service. The workbench does not call a provider or wait through real delays. Its configuration requires a maximum attempt count and a narrow status list. In actual code, consider whether the operation is safe to repeat and whether an idempotency mechanism is needed before retrying a mutation.
A worked artifact for a separate fictional task. Compare its actions with the contract.