I'll tell you what's interesting. This blog. It's from a fairly unsexy uptime monitoring company, whose services I've got little interest in. If the blog was just company updates ("we've got a new server" or "20% off all packages this week") it'd get very little traffic. But I've got it in my RSS reader and it hits the social news sites every other day with well-written and often humorous observations on tech.<p>Lots of people have heard of Pingdom who ordinarily wouldn't; it just goes to show how a little effort making your website something more than just a corporate website can go a long way.
One interesting fork was egcs, forked from gcc in the 1990s, by people who wanted to try out the "bazaar" development model (as opposed to the "cathedral" model the FSF used, which meant very little engaging of the community.) The experiment turned out very successful and eventually the egcs version was named the one and true gcc (and the gcc we use today is a derivative of it.)
Best line in the article, re: Carrier forking from Pidgin: "A fork (initially called Funpidgin) was done of Pidgin 2.4.0 because there were disagreements about the size of the text entry field."
This article is interesting, but I think it is too shallow to get any deep understanding of causes of forking as a phenomenon. A case study on /that/ would be kick ass, and quite a treat in terms of both software engineering methodology and oss anthropology.