As described in T279440#7489667, we have observed some drift between the primary analytics_meta database server (an-coord1001) and its failover replica (an-coord1002).
This may be limited to the superset_production database, but it would be useful to ascertain whether any other databases have been affected and as such, whether we can trust the veracity of the replica.
If it turns out that the drift is greater than we expect, we can rebuild the replica on an-coord1002.