A typical job for my proposed moderator, that: Demand an answer to that very question: Why exactly would it not work, or not work to a sufficient extent? Forget about proven combos and generic roles – look at the actual argument offered, and at the particular traits/skill sets of the players in question, and then undermine the thing.
For me the “proven” thing has always been a matter of personal preference, almost, when it comes to these drafts: You can choose to put a lot of emphasis on it – or you can choose to ignore it altogether (the latter being my natural inclination). Both are legitimate, as far as I'm concerned. But one thing should not be in doubt: It says nowhere that you have to field proven combos, or that you have to field players in roles they played in historically (to whatever degree).
The focus should be on what the manager claims will work/happen, based on his particular game plan. What is generically and/or historically true may be looked at to help undermine the arguments, but it can't be cited as some sort of irrefutable proof that a hypothetical/fantasy scenario would not unfold. The latter makes no sense whatsoever, and all it undermines is the very spirit of these little games: Use the evidence to demonstrate what sort of player X was, not to demonstrate what roles he played: The latter is irrelevant if the manager goes for an experimental set-up.