With Ethereum contracts the cost of validation is expensive, and for proof of work you need a function that is expensive to run, but cheap to verify (the inverse of SHA-256 is such a function).
The main concern against proof of stake is what you just described actually: it doesn't cost money to create, but as the value of staking is part of the cryptocurrency, a malicious party could get and stake of a lot of that cryptocurrency, create a leveraged short contract to protect itself, then reverse the transaction.
Another example would be using protein folding for cancer research for example, which sounds a noble cause, but it creates an incentive for drug companies to be miners, and could lead to even bigger centralization of mining than what is already happening.
