There's not been any planned downtime of the service as a whole.
The postgres replication to some of the readonly replicas serving crt.sh:443 had fallen significantly behind. This was due to an increased log entry ingestion rate (causing more data to need to be replicated) and very high load on the readonly replicas from user queries (causing replication activity to be throttled).
We've had to take each of the affected replicas out of service, one at a time, so that the replication could catch up.