This ended up in my IM today, and so far, I'm mystified.
What situation would cause a conflict object (one with a CNF: DN) to exist only on GCs, like a zombie. There are no writable replicas of the conflict object anywhere, and the AD environment, through large and very distributed, is also well run and monitored. The objects in this case are user objects.
The conflict object would appear if two DCs originated a create operation for an object with the same DN within the replication latency window. But what would cause it to become a zombie?
This is a weird one.
More data as it appears comes in...