What is a safe retry strategy when an integration receives rate-limit errors?

lrichardson

Consider a scheduled integration that receives a rate-limit response halfway through a batch. Restarting the entire job might duplicate records, while retrying every failed request immediately could keep the integration above the limit.

How would you combine checkpoints, retry delays, and duplicate protection in this situation? What should an administrator see while the job is waiting to retry? Assume the external API may provide retry guidance, but some requests can also fail for reasons that will not be fixed by waiting.

emorris

Follow the provider’s retry guidance when present, otherwise use bounded backoff with jitter for retryable failures. Save progress and make writes idempotent with stable record keys. Authentication and validation failures need separate handling. The admin view should show the last checkpoint, pending count, next retry, and the point at which someone needs to intervene.