Nothing at stake or other similar technique are based on quite a poor Longest Chain Determination algorytm. And the algorytm is also responsible for diminishing PoS contribution to security.

We could demand both larger stake and larger pow difficulty to resolve the conflict iff favor of one of the chains. With this method the attacker will need both 51% of hashrate and 51% of working stake to fork blockchain as he likes. It's not ideal but with such a logic stakeholders will be able to prevent 51% PoW attack and pools will be able to reverse the chain in case of nothing-at-stake.