analogous to bug 697277. Maybe instead of rewrite of the logic we just ad a notion of "source" and "destination" VM so it's clear where does it happen and if it is relevant? Not a solution but it would at least help...
let's go with comment #1 then
I cannot reproduce this issue (I never got the "Exit Message" audit log). By examining the code it does looks like this scenario may happened, the list that used for storing running VMs is not synchronized.
After talking with Omer, This may not be the problem since the VdsUpdateRuntimeInfo is running in a single thread... Aborting the patch. Yoshinori can you please reproduce the message and attach the Engine and VDSM logs.
it seems to be the same issue as in bug 965525 *** This bug has been marked as a duplicate of bug 965525 ***