There are critical queries in complex applications which absolutely require querying the master instead of the replica which might not contain the latest entries.
Please think about providing a way to read directly from the master instead of depending on a timeout or timeout loop to wait for the replica to update with latest values.
Timeout will not work when generating incremental IDs as well.