As far as the community management stuff implemented in dotLRN we'll have to take a look to see whether or not this can be generalized in a way that's useful for the toolkit at large. My reading of Ben's comment is that this portion of the dotLRN code would serve as a better starting point than the code that exists in OpenACS 4 today.
Certainly today's hardwired classes and clubs don't match the software project paradigm (and there is hardwiring in parts of the code) so it's not off-the-shelf a perfect fit for a sourceforge-ish project.
As to what goes back in the toolkit and what stays in the dotLRN tree ... the community will largely determine that.
I'm not in the least bit worried that vertical apps like dotLRN will fork the community or weaken OpenACS per se. Just the opposite will happen, IMO.