I disagree. Current network hash rate is ~ 25TH/s. BFL says they'll give you 1.5TH/s for $30k in a single box. You'd need only about 20 of them to gain > 50% of the hash rate, that's $600k. Six hundred thousand dollars! So let's suppose someone tries to set up this attack in the future when there are already, let's say, 10000 such boxes already running and connected. You'd still only need about $300 million (ignoring price reductions due to efficiencies of scale, and improved technology) - that'd still be small change to a central bank that's printing, what, $40 billion per month? Suppose there were 1 million such boxes running - that'd cost $30 billion, 8 months of QE3. Let me emphasize, the attack would have to be surreptitious so as not to raise suspicion. ASICs coming online would be the *perfect* moment to implement the attack because nobody will be worried at sudden massive increases in hashes-per-sec.

ASICs will be great for bitcoin if they become truly distributed, but it should be obvious to all that they'll never be as uniformly distributed as CPUs. Any argument on that point?

If civil servants get paid in a particular blockchain, then they'll use it. And if the police understand that their bitcoins become more valuable by beating on 'blackchain' exponents, then you can be sure they'll do just that.