I have increasing replication lag on the hot standby. I grep processes
on the stanby and see that WAL writer is in status "waiting". This
always happens after (auto)vacuuming of table used by problem query.
Once I kill problem query on the standby WAL segments are applied and
the replication lag is no longer exists.
Forgot to mention that max_standby_streaming_delay = 0 and
max_standby_archive_delay = 0.