They never said they were using rsync on their datasets. They're likely using ZFS snapshots.