Last Updated: June 8, 2026
Remote calls fail, hang, or slow down. What the caller does next often decides whether the issue stays small or becomes an outage.
Timeouts limit how long a caller waits. Retries decide whether to try again. Used well, they handle short-lived failures. Used badly, they overload an already struggling dependency.
This chapter covers timeout selection, safe retries, exponential backoff, jitter, and protections that prevent retries from making failures worse.