We're investigating an issue where threads in a small number of servers are unavailable.
We've identified a data issue with active threads in servers that opted into an experimental, opt-in feature. We've fixed the underlying issue and are working on data recovery for the affected threads.