either its an internal service issue, or the service which handles the stt is getting blasted with 500s from deepgram if it is actually down and is getting overwhelmed