Hmm. Actually, one thing I've been meaning to do is change the way it gets that IP for the client's log, because what is being printed there now is just the result of calling the old gethostbyname() API and printing the first address it gives back. Nowadays the client does have code in it to work out exactly which of the network interfaces is being used to talk to the server (and thus, what the most appropriate IP address to print in the log is) but it's not yet hooked up to the printing code.
So, what you're seeing in the log is in one sense just a cosmetic bug, and that log excerpt seems to be showing that the main client code per se is fine. There obviously is a real problem but I think it's lurking a little deeper, and my intuition is that it's the configuration collection code (which lives in a plugin file called
MACHCONF.DLL that both the client and server use) that has the real problem; it may be having trouble trying to read the settings for the VPN adapter for some reason.
What version of the Nortel VPN client are you using (I'm assuming that the client OS is XPSP2, by the way)? When I last used their Extranet Access Client some years ago, I recall that it used to properly disable the VPN adapter rather than leave it active but fall into the Ethernet autoconfiguration system, which is what is happening for you. It may be that there's some design changes in different versions of the VPN code to match the changes in different Windows editions, and I'd like to make sure that we use the same version to reproduce your situation more exactly.
If you could e-mail me the full client log that'd be helpful, too. In the meantime we'll try and get hold of various EAC client versions and experiment with them.