For a period of approximately 90 minutes from 0219 IST(GMT +530) to 0357 IST(GMT +530), we were experiencing timeouts and high response times for API requests on new uncached image transformations, the `/v1/metadata` and `/v1/files` endpoints. For image delivery, 0.3% of the requests during this interval were terminating in an error.
This was a result of one of the shared servers in the US East region becoming slow and eventually unresponsive in this interval. Eventually, this server had to be taken out of the group manually, which helped to restore the services. Our team is working on upgrading the infrastructure in this region and improving our automatic server isolation mechanisms to prevent such incidents from occurring in the future.