1) brake replication relationship occasionally (which is not the best idea), and bring your volumes online in this way, and then reestablish relationships & continue replica. But in this way along the side from data testing, you'll test your DR behavior like in the real world.
2) or make SnapVault to third storage or the same DR storage and make FlexClone from each (or required) volumes, now you can test them either with the same subnet or different. In this way, you will have basically 3 data sets: one on primary, one in SVM-DR and one in SnapVoult relations. Clones don't need to be split.