Currently the broker does not become active until all persistent queues have been fully recovered. In the case where queues have a large number of entries this may take considerable time.
By adding a mechanism by which the queue recovery can be done in the background while the broker becomes operational, the return to operation time of the broker can be improved.
Newly published messages should not "overtake" messages already in the queue (save for already permitted exceptions, such as priority queues).
Note that int he case of priority queues, if consumers attach to the queue before the queue is fully recovered older but lower priority messages may be delivered before newer, higher priority, messages