Retry and Resilience Logic

Help me add retry and resilience logic for calls to {{dependency}} that sometimes fail transiently. Recommend which errors are worth retrying versus not (do not retry a 400, do retry a timeout), exponential backoff with jitter, a retry limit, idempotency so retries are safe, a circuit breaker to stop hammering a failing service, timeouts, and a fallback. Provide the approach. Note how naive retries can amplify an outage and cause cascading failures, and how to retry responsibly. Stack: {{stack}}.