Faulty rate limit tracking

Yes, we set the rate to 100 per second for an API and ran a load test. They say they didn’t see more than 5-6 requests per second at any point. Still, they observed failures due to 429 errors.