When running any other type of operation block in a loop, when I receive rate limit or other errors, the block recognizes it as an error and proceeds to the next step for error handling in the workflow. However, it appears that AI workflow blocks don't recognize these types of errors and instead processes 429 responses as if they were successful responses.
For example, in my case, I make GPT4 calls looping through data and collecting the responses in an array of response objects. Some of those response objects are 429 rate limit error responses, which I would not expect because I have the block configured to retry with a delay per iteration if the iteration fails.
Can you share a screenshot of your block timeout and retry settings.