A document flow handles a few files successfully, then starts returning 429 errors when a large batch arrives. More immediate retries can add pressure to the same service. Check the failed action and control the incoming work so files either complete or remain visibly queued for review.

Identify the action that is being throttled

Inspect the run history and record the connector, response code and time of failure. A 429 response indicates too many requests. The limiting action may be the AI processing step or a later write into another system. Microsoft’s Power Automate throttling guidance explains how connector and platform limits can affect flows. Verify the applicable limit for the actual connector and configuration.

Bound the retries and the work in flight

Check whether the connector action already retries. Avoid placing a fast retry loop around an action that has its own retry policy. Configure supported retries with increasing delays, honour a service-provided wait instruction where the integration exposes it, and cap the attempts or elapsed time. Microsoft’s flow error handling guide describes retry policies and failure handling.

For a predictable batch, process fewer files concurrently and measure the resulting completion time. If arrivals routinely exceed processing capacity, persist a queue with a document reference, attempt count and status. Queue records should survive a failed run. A file that reaches the retry limit needs an assigned review path and enough information for staff to resume it.

Test recovery without repeating the business action

  1. Use a controlled test response or connector sandbox to exercise the 429 path. Observe how long the flow waits and where it ends after exhausting attempts.
  2. Check that a document has one status record and that retries reuse its reference.
  3. Verify the destination before replaying a failed run. A write may have completed even if the response was lost. The failed-write review guide covers that exception.

ITZ’s Power Automate workflow service can review batch intake, connector retries and exception ownership. Describe the failed action and your usual batch size.

Review your document flow retry plan

Bring a redacted run history and the required processing window. Review Your Document Flow Retry Plan