Maybe you can provide us with some of this miracle code, then? We're always looking for good patches...
Maven 2.1.0 did not use the httpclient-driven wagon; it uses the Sun implementation. Also, Artifactory may have different licensing requirements than Maven, which may allow them to use code derived from jCIFS that is known to support NTLMv2 according to some other threads I've read.
A big part of the problem here is that I don't have access to an NTLMv2 proxy for testing, nor would I really know whether/how I could set one up on a VirtualBox/VMWare appliance for testing. This is the root problem when dealing with proxy issues, since there are a million flavors and they all seem to have different quirks. Are you sure you're behind an NTLM v2 proxy, or is it just v1?
We're just using httpclient, there are no tricks or hacks. We have a client instance initialized inside the HttpWagon (actually, AbstractHttpWagon, IIRC) and it's used in the same manner as I've ever seen httpclient used anywhere. There's nothing special about it, but we may not have exactly the same configuration code as you do. So, I could use a look at your working example, to compare notes.
BTW, I'll start trying to release Maven 2.2.1 as soon as our issue count hits zero for this version. As time goes on, we may shift some issues out to 2.2.2 or 2.3, and we may discover some new regressions that need addressing before the release. But in any case, I'm working as hard as I can to get this release done ASAP.
As always, users can help out immensely by providing sample code and failing test cases that can be used to create integration tests.