The connector's streaming loop tracks what the current redo log is by name. In the event that enough log switches occur that Oracle wraps around to the same active redo log but the logs now have different sequences, this would mean the loop's log switch check would fail and likely will lead to a connector failure.

To avoid this problem, the check should track current log file changes by sequence number rather than by name as sequences are unique to each instance of a log file.