I always see people quote the Gavin phrase of "nothing at stake" as fact, but it seems like a complete logical fallacy to me.  Don't you need a majority of people holding the proof of stake currency to run non-reference clients to stake multiple chains at once for this to be true?  So you're basically saying the people with the most financial incentive to protect the network are all going to attempt to destroy the network security on purpose, and do it en masse?  How does this not violate game theory?

Bitcoin relies on game theory to exist.  It seems like these same principles should prevent such a thing occurring in proof of stake.

Let's not forget that most people on the planet aren't programmers.  If any crypto coins make it into the mainstream, the majority of people holding proof of stake coins, or using the network at all, will not have any idea how to leverage their stake to attack it.  I feel like many proof of stake attack vectors are done under the assumption that the majority of coin holders are Russian cyber criminals.

Assuming the proof of stake complaints are either dispelled or fixed, the main issue I have with proof of stake is it's attacks are very hard to see coming.  If a vendor like Amazon was taking huge volume of Bitcoin, they would actively monitor network hashrate and have a Defcon 1, red alarm siren go off when the hash rate became too lopsided to accept BTC.  They would then just not accept payments until the hash is distributed better.  Practices like this make the 51% attack almost irrelevant for big business.

Doing something similar to the above with proof of stake seems almost impossible.