This setting only comes into play when a thread makes a full top to bottom pass on the queue records it processes. Polling every two seconds simply means it restarts the process again very quickly. While this might seem a good idea, during periods of low queue volumes the thread processes are constantly attempting to do reads and write logs. A typical value would be 60 sec sleep interval. if the table/rows in question are not moving quickly enough for this use-case, moving this table into a thread with no other tables might be a better practice, or into a thread with lower overall volume than the thread onto which it is currently assigned. Queue volume by thread is the main factor in events not being replicated in a timely fashion. If the table is in a thread to it's own, then volume lags are typically eliminated.