Resolved. Between about 06:40 and 08:00 UTC a portion of requests to the API and the website returned errors, in two short bursts — the heaviest few minutes were around 07:15, and there was a second, briefer burst just before 08:00.
Affected requests failed outright rather than returning wrong answers, so there is nothing to re-check: anything that received a response was correct. Most requests during the window still succeeded.
A small number of submissions failed with an error and were not saved. If you sent a submission in that window and got an error back, please send it again. Everything that was accepted is queued and is being processed normally.
No stored data was lost, and nothing already in the database was affected. Backups continued and have been verified since.
The cause was a failure on one of the machines running the service, which for a few minutes disrupted the internal lookups that let the different parts of the service find each other. Databases moved to other machines automatically, which is what caused the second, shorter burst. Service recovered on its own each time, and everything has been running normally since 08:00.
Public data file downloads at data.acoustid.org were unaffected throughout.