Extremely sorry about the downtime. Scaling up to multiple machines has revealed a peculiar issue with a scheduled process that's supposed to run every night... once. It was run on every machine instead, putting the database in a deadlock state.
We're back now and I'm going to add a fix before next night 😩