You can configure a route from one Replication Server to another, or a connection from a Replication Server to a database, so that the Replication Server stores stable queue messages for a period of time after delivery to the destination. This period of time is called the save interval.
The save interval creates a backlog of messages at the source, which allows the replication system to tolerate a partition failure if it is corrected within the save interval. If the stable queues at the destination fail, you can rebuild them and have the source Replication Server resend the backlogged messages.
Refer to the Replication Server Administration Guide Volume 2 for details.