4 sites, connected by fully meshed VPN over T1s. Each site had one 2003 DC.
About 18 months ago a DC failed. It was rebuilt with the same name, and promoted to DC, however it didn't function properly. I wasn't there at the time so I don't have any more details, but they ended up with a new server with a different name configured as the DC in that site.
I don't know how the failed DC was demoted, but I assume it wasn't correctly.
GPOs only work in one site. The other site's DCs still share SYSVOL, but the data is now over 18 months out of date.
Steps taken: I've run FRSDiag.exe, but most of the output is beyond me. When I was selecting servers it offered to scan the original failed DC. This makes me think the FRS configuration wasn't updated when the failed DC was demoted, but I really don't know.
The summary report for each server from FRSDiag can be found here: http://pastebin.com/f458b50fb
Event logs from the servers: http://cid-fa226a18f1d8248d.skydrive.live.com/self.aspx/.Public/Johnsburg%20logs/Johnsburg%20Logs%20all.zip
For reference the active DCs are: HSServer2, JServer, RServer, BServer1.
I found this TechNet article, but I was hoping for a more subtle approach. http://support.microsoft.com/kb/315457