What should happen when a provider says slow down?

A connection that works for one record can fail during a busy import. Providers may limit request rates. Repeating the same call immediately can make the problem worse and leave staff unsure which records have been processed.

Read the actual response

The HTTP 429 reference explains a standard rate-limit response and the possible Retry-After header. Your provider's current documentation determines its exact limits and recovery rules. Different accounts may have different allowances.

Keep work identifiable

A Colombian supplier synchronizing a catalog should be able to see which batch is waiting and which products completed. Stable identifiers and progress records help a resumed job continue without duplicating or overwriting unrelated work.

Set a stopping rule

Retries need a limit and an escalation path. A missing permission, invalid record or closed account is not necessarily a temporary capacity problem. Staff should receive an actionable description without secrets or unnecessary customer data.

Test with a small batch and simulated provider failures before a large migration. Agree acceptable delay and who investigates unresolved items. A well-scoped integration includes operational visibility, not simply the code that sends the successful request.