I like to think I'm friends with a bunch of IM developers, and I'm well aware of their opinions of Jabber. If you feel compelled to bitch at me, you may do so on this post. (Know in advance that I don't especially care, nor do I have any influence on this project. Nyah.)
But protocol snobbishness aside, it seems to me to be an obvious choice: it's not creating the n+1th protocol, you get built-in support from every interesting IM client, and you get library support (there are e.g., Ruby Jabber bindings, but not Ruby OSCAR). One good way to look at it is that it's the least bad of the possible options.
All this "client choice" talk surely reads like marketspeak from a company that doesn't have market share yet, but I also think that if you think about it, lock-in is relatively uninteresting to a company like Google (especially as compared to AOL and Y!), as we typically shine when our products can be compared head-to-head. (But again, I'd like to point you at my sidebar.)
For more reading: the discussion here is worth taking a look at.