In the last few days, we've have 3 instances of wdqs-updater lagging behind on some nodes (but not all). It looks like replication stopped for some time and then restarted and took some time to catch up.
The fact that not all nodes stopped at the same time indicates that the problem is internal to wdqs.