For 2 operating sync servers, you'd rely on very robust recon services on all of the target systems. It gets really complicated when those target systems have active-active clusters, and no write masters.
For this customer, both data sync servers in separate DC's will be configured to write to its local Sun DS instance. Ideally, 1 sync service is running, so data flows in 1 direction.
What they are concerned about is that, after connectivity issues are resolved, the 2 Sun DS instances will start replicating again. If new users were created during this time, both Sun DS instances will have the same user records, but all with different entryUUID values. This would result in replication conflicts which they would have to address, it has caused them grief in the past as they weren't sure which one was the best one to keep.
They also only have 2 Sun DS instances, so they don't have a majority rules setup.