Resolved -
A recent production regression in our tts/bytes endpoint allowed certain malformed requests to queue indefinitely, causing affected accounts to hang. We identified the root cause and deployed a fix by 2:45 PM PDT, fully restoring service. The impact was limited to a small subset of users. Going forward, we’ll tighten request validation and strengthen our deployment safeguards to prevent recurrence.
May 1, 13:00 UTC