Thankfully the problem that I was having (very similar to your) has not returned. I was able to unwind all the VM by DR copying them to a new SR and then deleting the originals.
I am really interested in knowing what has caused yours to return because it has the possibility on my stack as well.
Only thing I can tell you about my setup is that I am doing incremental backups every night on about 200 VM's. Some of the VM's I replicate XSite, but I only do that on the weekend and I make sure that neither job can run at the same time.
What I have not done is a full backup on the VM's as a separate job. I have my full backup interval set to 30, so roughly once a month it will cause a full backup vrs incremental. I also know that some of the VM's have already done this and the problem did not return.