Inference API is currently experiencing a High 502 failure rate. Roughly 45% of all requests are affected. Our team is currently investigating the cause of the issue as a matter of high urgency.