Hello,
I had a similar issue before. We had SRM in the mix and because you cannot storage vMotion a VM with replication enabled, it ended up having a split brain scenario after we tried to move it to a new datastore. This sounds similar in nature to what you are seeing if I understand correctly.
You need to look at the location of the .VMX file because it is likely spanned (in two different places) across the two datastores. We found that there was a zombie VM (simliar to yours) on the same host showing no cpu or memory utilization. The only fix was to reboot the host, remove the VMs from inventory and then move the .VMX file back to the correct location (same datastore directory) or point the VMX file that was on the destination datastore back to the source .VMDK directory. Then add that VMX back into inventory.
I hope that makes sense and is helpful!
Cheers,
Tom