It depends heavily on the journal. Most journals have a "final draft policy": What they print is only theirs to publish. But you can self-post whatever previous versions you have. In my case, I think there are one or two minor spelling mistakes in the versions I have posted in ArXiV and my homepage. It does not take extra time (at least not a lot) to self-post it or publish on ArXiV (just a little hassle with image conversion problems, YMMV)

Would you say that you're the exception in posting them for the public? If so, would it be worth while for someone to try to get at these non-final but still perfectly useful papers?

I really can't tell. As far as I know, all people in my department publish freely his documents: either in ArXiV or in the department page for submitted papers. Also, ArXiV has a huge numnber of articles in Mathematics, the growing trend is to submit it there. I guess that most mathematicians (or at least, young ones) at least provide some draft version of their published manuscripts online, freely available.

Yea, ArXiV is a great resource, and I actually hadn't seen that it's grown this much. Your department is one of the good ones - I salute you! Here's to more doing the same.

I also hope everyone starts doing it. There is no point in making research unavailable to the public just for the sake of keeping the journal's "level". The future is open content, but most publishers are still blind to it