At least some of Haze's more recent posts (like "_and it's all pointless anyway?") have more accurately represented the proposed changes, which he clearly didn't have complete information on when he started blogging about them. Personally, I think a MAME with core files BSD and driver files LGPL would be a good thing, and mutually beneficial for the emulators incorporating each other's code. However, it sounds like MAMEdev is unable or at least unwilling to relicense code without the unanimous consent of the primary contributors.
So I guess the short-term goal is just to tag various files with more permissive licenses if they're able to do so. I don't see a legal problem here as long as they're making a good-faith effort to attribute copyright to the legal holders, and correct mistakes as they're brought to their attention. Simply contributing technical information, as useful as that is, does not grant copyright to the resulting code others write. And while there will no doubt be gray areas, there are legal guidelines as to what sort of code changes are copyrightable.
But I am curious about the long-term goal. Are you trying to eventually get to an OSI-licensed MAME? Perhaps this is possible long-term if you get agreement from a large number of contributors, re-write a bunch of code where a copyright holder doesn't agree to the change or can't be located, and don't accept submissions that would change a dual-licensed file back to a MAME license only. That would of course be a lot of work that didn't immediately improve emulation, but maybe it would be worth it in the long term. If you're not trying to do that, then I guess you just continue to have MAME with scattered files tagged with secondary licenses. That might be enough to benefit some particular drivers and projects, but I don't see it helping with some of the other uses discussed like museums or academia.