Possibly vitess from the latest update:

> primary failover briefly improved performance but did not fully mitigate, we've throttled inbound traffic and are investigating upstream Vitess issues

Thanks!