True, and as the value of those potentially doubly-spent coins increases, so does the cost of preventing such an infraction. This is necessary to protect the at-risk merchants whose presence on the network is vital if BC is to succeed. Also, massive loss of confidence resulting in destruction of the value of my assets would have the same net effect to my wallet as theft. So, it is in our general interest that BC transactions by required to pay into a commonwealth of miners who will burn GPU for us. This should be taken out of the now-optional transaction fee. Satoshi apparently had anticipated this. Under such circumstances, I believe your models of power consumption are heavily understated.

Given that there's nobody who can force any particular action, it'd be interesting to think through the game theory and whether that's actually likely to happen.

An assumption I have made is that most merchants in the "normals" category would demand near-as-dammit watertight protection from theoretical yet systemic risks. FUD from competing systems would drive this requirement. I cannot imagine a lab experiment to test my assumption; it would need to be demonstrated in the marketplace. The cost of providing a very strong defence is entirely disproportionate to cost of offence, so we are looking at an asymmetric arms race.

Many Game Theory experiments are concerned with discerning underlying morality in economic exchanges. An interesting perspective here is that there should be no room for morality when Bitcoin gets up to scale. (But thanks again for the good guys who saved the day this week). No black/white hats; only those who have been paid to protect and those whose interests lie in exploitation. Eventually no actor will have the resources to do the right thing, unless it is also explicitly in their short term material interests. That ethos seems to be prevalent in the Bitcoin and one of the reasons that I find it interesting to watch.