I wanted to check if anyone has faced this issue with Azure Open AI. I was doing some embeddings and suddenly started getting 429 errors. I was not hitting the API hard, the requests were minimal. So, it seems that the rate limiting probably is not at the user level, it may be at all user(who are using the service) level.
Just thought to check if anyone can confirm this. For now, I implemented an exponential back-off to address this, but for the limited number of requests, I should not have got 429