I'm seeing some strange behavior occasionally when cloning from template. Sometimes shortly after powering on for the first time, DRS moves a VM from one host to another because it's lost heartbeat. If this happens while guest customization is still going on the customization complete event is never written in tasks & events. This would not be a problem normally, but vRA appears to key on this event to determine that customization has finished, and without it the request just gets stuck in the CustomizeOS status.
I've compared the guestcust.log between machines that finish normally and ones that didn't, and found a couple of discrepancies. In a "good" build I see this:
[2015-10-23T10:11:18 DEBUG] Rpci: Sent request='deployPkg.update.state 5 0 C:\Windows\TEMP\vmware-imc\guestcust.log', reply='', len=0, status=1
vs this in a "bad" one:
[2015-10-23T10:11:31 DEBUG] Rpci: Sent request='deployPkg.update.state 5 0 C:\Windows\TEMP\vmware-imc\guestcust.log', reply='Unknown command', len=15, status=1
The bad one also has this:
[2015-10-23T10:11:31 ERROR] Unable to set customization status in vmx.
That last one seems pretty damning, but I don't understand what's different about these builds that causes that communication to fail. Couldn't find anything pertinent in the KB either. Any ideas?