I've read this thread ago, when it comes to NXT implementation of PoS, arguments are simply wrong.

Let's assume you have a lot of NXT. The fact that you can forge block with n traansactions with score X, doesn't mean that at the same time you can forge block B with similar score.
So yes in case of PoS you could SEARCH{*} for a block that would give you best score, but that doesn't mean it is beneficial for you to run "multiple forks simultanously".

{*} by including or excluding some of the TXes

People seem to simplify PoS to "bigger stake -> better 'mining'", when that's actually not the case.