The .0.12 GB for replication is used for storing the metadata, block changes.... not for cache. There is no setting for cache. If connection is lost data that has not been transferred during the lost connection will sync once connection is re-established then the synchronization of the data from the source will continue to destination as it should be.