Sadly, it's not that easy. Removing the need for centralized trackers (which would basically make the protocol no longer BitTorrent in any meaningful way; it would really have more in common with Overnet or Soulseek or something) just moves the bar down from trackers to major seeders, whose IP addresses wouldn't take long to collect. Bram Cohen has said from the beginning that BitTorrent was never intended to be a "safe" protocol for warez or whatever, it was intended to alleviate server load for large legal downloads (Linux ISOs, amateur films and the like), and it's extremely good at that.
Freenet has been working on a "safe" network for years, and while the concepts are sound on paper, it relies on emergent behavior to allow knowledge-deniable routing, and the reliability of the network has been spotty at the best of times due to subtle and not-so-subtle emergent flaws in the routing details (GNUNet, Entropy, MUTE, and antz are based on similar concepts, but different implementations; Freenet was the first and AFAIK the largest and has barreled headlong into more bleeding-edge problems with the concept than any other system to the best of my knowledge, which is why I mention it in particular). A lot of P2P folks are just now getting around to looking at problems that Freenet was designed from the beginning to avoid (the stated purpose of Freenet is to avoid rogue or oppressive government agencies, so they didn't wait for the MPAA to start busting heads) but that others considered to be too paranoid / anti-performance. But you can't get real protection without paying a price in performance. Measures like PeerGuardian only stop the most obvious of attacks and would not be difficult to counter for an organization that is serious about attacking a network. If you really want to get a picture of how hard the problem is, I recommend reading the old freenet-tech mailing list archives where various potential attacks were discussed. Truly anonymous/untraceable filesharing is like the cold fusion of P2P - every so often someone claims to have done it, but they seem to invariably run into problems that have been known for years by more experienced researchers.