At the same time, the value of solving large thefts like this makes expending substantial resources to solve it much more reasonable.
It's like stealing someone's iPhone with a ski mask on while a camera catches you vs. robbing a bank with a ski mask on with a camera catching you. Both thefts begin with the same amount of evidence, but the latter is infinitely more likely to have resources devoted to it.
If a drug buyer in a BTC transaction got the coins anonymously (say, sent a self-destructing robot to drop off cash in exchange for BTC), then that part is fine. They can go use those coins anonymously. If they ship drugs to their house, well... that alone sort of undoes it all.
The drug sellers that were caught from Silk Road were caught due to things like making a huge amount of trips to post offices, getting the attention of postal workers. That may be a parallel construction, but it seems legitimate enough. Buy drugs from large seller, look at the postmarks for patterns. Then go gumshoeing around and wait 'til you see the same car or people going to the same post offices over and over and over. You can probably pull this attack off without even having government capabilities.
I had never really thought about the consequences of a public blockchain and the decreasing cost of analyzing that blockchain.
I seem to recall a news story about how people who spied against US in the 40's and 50's were caught later in the 70's and 80's because the intercepted cipher texts had been kept for decades until technology was able to break and decrypt them. Looking at the blockchain reminds me of that feat. Of course, that could have also been a poorly written thriller I'm remembering instead...
If you use it correctly, then bitcoin itself is anonymous.
a) Bitcoin is pseudonymous enough that you can safely buy drugs with it
b) Large thefts of bitcoin can be traced; it's not pseudonymous enough for massive frauds