While we're speculating, I'd like to say that some other things that would be very interesting to look at are:

1) Active attacks on anonymity, on the bitcion network.
There's some people using mixers.  But how do you know your coins are really mixed?
Lets say you trust the mixer.

But what if your coin is mixed with a bunch of other coins, all of which belong to an adversary?
If I was interested in actively attacking Bitcoin, I'd be flooding mixers all the time.

I could make it appear to another user that their coins were mixed, when in actual fact, I controlled all of the coins they were mixed with, and could tell for sure what the incoming and outgoing coins were.
Obviously, as the mixer takes a fee, there's a cost, in Bitcoins, to doing this.

But, while I've seen a lot of talk on mixers out there, I haven't seen this sort of threat mentioned (maybe I'm missing something - this is something to consider, not something I've thought about in depth.

2) The IP layer work that Dan Kaminsky did - could that be put together with Bitcoin layer work like we did?

3) There's whole classes of timing and statistical attacks we didn't consider.