HEARD: The checklist for not getting your integration rate-limited into oblivion.

HEARD: The checklist for not getting your integration rate-limited into oblivion.
Per an engineer who maintains a dozen platform connectors, most 429 storms are self-inflicted. Their defense:
— Read the Retry-After header and obey it — don't guess
— Batch reads; never loop one-record-at-a-time when a bulk endpoint exists
— Add jitter to retries so your whole fleet doesn't retry in lockstep
— Cache reference data (lists, pipelines, owners) for the session, not per call
— Track your remaining quota header and throttle yourself at 80%, not 100%
The nastiest outages come from a retry loop that turns one 429 into a thousand. Watch this.