Standby masters appear to not purge their old write-ahead logs

Dropping this one here, haven't investigated yet. We noticed that on a cluster who's elected master had been active for quite a long time that the master store wal directory steadily grew in size. Failing over master cause the directory to be cleaned up.