This is a great theory, but still doesn't explain why it haven't happened in reality. You can make up scenarios and theories all day of "what may happen", but when it doesn't actually happen in reality, you have to question your theory, not question reality.
Gavin already addressed your question about why it hasn't already happened in reality. The fact that an attack has yet to occur historically is no way to design a secure protocol either. The fact that you are even using this as an argument shows you aren't serious about security or PoS flaws.
Yes, in Network and Information security one must analyze and prepare for all hypothetical attack vectors. This is the reason most Bitcoin proponents who understand Information Security still suggest Bitcoin is "fragile", 7k average active nodes is too few, and how they would prefer other implementations like libbitcoin to exist alongside Bitcoincore that interact with the blockchain.
Gaven's response was basically "they aren't big enough yet". I can't really see the logic behind "being bigger makes it easier to attack". Tiny PoW altcoins gets 51% attacked all the time, even sizable PoW like Dogecoin gets 51% attacked several times, while the biggest ones Bitcoin and Litecoin never got attacked.
If anything, I would think being bigger makes it much more difficult to attack.