Here are a few more papers from SIGCOMM which should be of particular interest to a more theoretical audience. (Generally, SIGCOMM papers are interesting -- but again, I'm focusing here on papers that I think might be of special interest to theory people. It strikes me that I should, at some point, similarly summarize papers from a major theory conference -- like STOC 2009 -- that would be of special interest to networking people. Of course, SIGCOMM makes that easier, posting abstracts and all the papers online...)
There's a paper on analyzing BitTorrent in the game-theoretic incentive-style analysis sense. It will require a more careful reading from me, as I'm not a full-fledged game-theory/CS type researcher, but it sure looks interesting on the first perusal. I'm naturally biased to the idea that if all this current effort on game theory that is going on in computer science (and particularly in theory) is to have payoff, real-world protocols must be considered and analyzed. So in that sense, this should be a really interesting paper.
While it doesn't appear particularly theoretical (it looks like what I like to joke is a standard networking paper -- lots of pictures and tables, no equations...) this paper on spamming botnets from Microsoft includes Rina Panigrahy (well known for his work in both theory and practice) as one of the co-authors. (I figure Rina had something to do with where I saw the words "entropy reduction", but that's just a guess...)