Retry
Re-execution of a failed operation under an explicit policy for eligible failures, delay, attempt limits and idempotency.
Retry Only Recoverable Failures
A retry is useful when a failure is transient and the operation can safely be attempted again. Permanent validation errors, authorization failures or deterministic bugs should not be retried blindly.
Exponential backoff and jitter reduce synchronized retry bursts.
Correctness Boundary
Repeating a write can duplicate side effects unless the operation is idempotent or protected by a deduplication mechanism. A retry policy therefore belongs to the business/transaction semantics, not only the HTTP client.
Retry storms can worsen an overloaded dependency, so attempt limits and backpressure matter.