A retry strategy that doubles the wait time after each failure, the standard pattern for graceful retry under partial failure.
Exponential backoff is a retry strategy where the wait time between retries doubles (or grows exponentially) on each failure: 1s, 2s, 4s, 8s, 16s, often capped at a maximum and combined with jitter. The technique gives a struggling downstream room to recover instead of being hammered by a tight retry loop. AWS SDKs, gRPC clients, Envoy retry policies all default to exponential backoff with jitter.
A naive retry-immediately loop turns a transient downstream blip into a self-inflicted DDoS: every client retries simultaneously, the downstream gets crushed by retries on top of original load, and recovery takes longer than the original failure would have. Exponential backoff converts the retry storm into a controlled queue that the downstream can clear at its own pace.
See the part of the platform that handles exponential backoff in production.