Jeff has more thoughts on program-committeeing: his main point being that with the number of back-channels that already exist to favor the well-connected in the review process, adding one more (PC members contacting authors for clarifications) isn't that bad.
In a sense, he is right that backchannels always exist and always will. After all, any conference review system is an inherently error-prone process of determining quality/correctness in a short period of time. Precisely because of this lack of time, it is impossible to evaluate a paper in vacuo, solely on its stated merits, and thus extraneous factors like authorship will play a role.
What interests me though is a related point. He mentions the back channel of going around giving talks about one's work. The reason that this is an effective back channel is that it helps a potential reviewer: after all, if they hear my one-hour talk, they will get a much better (and possibly much more positive) view of my paper than they would have acquired merely by reading it.
In that sense, good marketing appears to be an important part of the process. In fact, in the past people have recommended that I go and give talks on my (unpublished) work precisely for this reason. However this goes back to a point I was discussing earlier: in a double blind review process, it is generally felt that one is ethically bound not to do anything that might break anonymity (give talks, put papers on web sites etc). But is giving a talk on a paper under review (in a non-double-blind scenario) not similar ? After all, doing this gives me a competitive edge on the poor soul from Outer Timbuktu who can't travel to important places giving talks, and sending out drafts is fraught with its own dangers.