The authors don't detail much experimental evidence about their new taint model (basically FIFO accounting for taint), but they do offer a couple of numbers which, if supported by more testing, would be quite impressive. At any rate, there's ample food for thought.
If the legislative landscape they hint at really comes into being, I don't know where I'd stand on the line between "hard" and "soft" approaches. Bitcoin enthusiasts? Investors? Law enforcement? All of these positions would entail both "good" and "bad" practical consequences from my self-interested point of view.
Nice piece of work. Computer scientists discussing legal implications with a practical, technically sound background. A rarity.
Tasty find merited.
One of the problems with such a 'taint' system that it doesn't address, is what of coins falsely declared stolen? In the paper they seem to assume all such declarations to be honest, but in practice it just moves the arguments to the lawyer-enriching 'how to get these coins declared stolen or not' level.