Incrementing txn_id from 0 is not a great idea, and we should be retrying messages rather than ignoring them. Investigate options around queuing