I have about 15 servers in different locations, all except 1 (an NT4 box that runs an app we can't upgrade) is w2k3 and is both a GC, DC. The I have had problems with replication and by repeatedly cleared all the automatic connections, rechecked topology, rebooting have been able to get the replication errors corrected. But the problem reappears after several days. The bridgehead for the AD is at Corp HQ and so all other DC's in the satellite offices 'should' replicate directly with that server. When things are working the automatic NTDS connections in sites and services correctly point to the CorpHQ server, but once problems develop, multiple automatic connection objects are created to multiple sites to try to replicate but they all fail. The symptom is that RPC communication fails and issues such as logging /mapping drivers to the satellite server is affected in addition to the obvious AD issues so if left it eventually directly affects users. Communication between sites is via a Cisco hardware VPN (most sites have t1's) and I have done various tests to show that there is no sign of any other sort of connection problem when this occurs - it appears to be a Windows issue completely.