Retry
Try-again strategy on transient errors (network timeout, 503). Default in modern SDKs. Must be paired with exponential backoff + jitter to avoid overloading the server.
Exponential backoff: 1s, 2s, 4s, 8s… Jitter: add random variation to avoid the thundering herd when many clients retry simultaneously. AWS SDK + Google Cloud SDK + Stripe libraries have built-in retry policies. Retry only on idempotent operations (otherwise duplicate posts). Circuit breaker pattern complements: after X failures, stop retrying for a while. Often combined with dead letter queues.