I, too, am having lots of trouble with "ghost" disks, and only since V8 Patch 2. In my case I'm replicating 2 VMs across a 50Mb fiber wan... a configuration that has worked well for a couple of years. But after upgrading to patch2 I can't get this particular job stable.
It starts with the replication job failing (after several nights of good replications) with "Error: Failed to open VDDK disk [[datastore1] arena1_replica_mid/arena1-000007.vmdk] ( is read-only mode - [false] ) Failed to open virtual disk Logon attempt with parameters [VC/ESX: [vcentervm.mydomain.com];Port: 443;Login: [root@localos];VMX Spec: [moref=vm-19505];Snapshot mor: [snapshot-20408];Transports: [hotadd:nbd];Read Only: [false]] failed because of the following errors: Failed to open disk for write. Failed to download disk. Reconnectable protocol device was closed. Failed to upload disk."
And then subsequent attempts of the same job fail with: "Processing arena1 Error: Detected an invalid snapshot configuration. Processing Gringotts-ACS Error: A general system error occurred: Failed to lock the file" and when I look at the proxy VM at the replica site it has both drives for arena1 and Gringotts-ACS still mounted and I have to disconnect them from the proxy manually.
If I delete everything on the replication datastore and start the job fresh again, it works for a few nights and then the same loop happens.
YES, I will open a case with support, but wanted to put this here as it seems very similar to the OPs issue. Something about patch2 changed the way my replication jobs work over the wan, and they ain't happy.