during a testfailover with VR we basically attach a redo log disk(s) of the current base disk in a new sub folder in the target datastore, i guess this is VR's version of a snapshot that the array based SRA's initiate on the lun if you want to think of it like that.
when you end the test we simply power off the VM, remove the subfolder (and redo disk) and we're done. you can now run another test if needed and at no point was replication interrupted. All other UI actions pause/remove replication do NOT relate to the test VM image they relate to the actual protected VM. If you are saying you want to run a test failover but then split off a copy of the VM to keep it at the DR site BUT also keep the original replication in place then during the test failover once the VM's are powered on in "test" mode you will need to make a clone of them as Martin suggested. if you end the test and hit "remove replication" it will unconfigure replication and then delete the disks that exist at the recovery site for that VM.
as a side note the following blog explains the difference during normal operation for using "pause" versus "stop" or "remove" replication:
Stop vs Pause with vSphere Replication | VMware vSphere Blog - VMware Blogs