The core argument is that there is no objectively determined network. A node that was not around during the time the "honest network" progressed has no basis of knowledge for which fork to choose when presented with equally valid options. In this case, "making up 3 months" is as simple as creating the blocks near instantly with only a signature as proof and no immediate cost. With PoW this immediate cost is very high for bitcoin, but can drop dramatically for many altcoins.

However, the argument started as a criticism of NXT and Peercoin where there is literally no downside to staking several competing forks. It has been reformulated several times over to apply to any proof of stake system (including ones that punish bad behavior)--somewhat successfully in my opinion, but only given some highly implausible (but not impossible) conditions. There is *a lot* of manipulation in the cryptocurrency sphere, so discounting implausible scenarios as impossible seems like a logical mistake. However, I think the future of cryptocurrency security will be in currencies that are more PoS-like than PoW-like.