Even with constraints on research, there is still the issue that players will always analyze the pure stats of each item and then just build the best (or at least the combination that gives the best bang for buck). For example, a player would just build P51 as the standard US fighter, yet real issues means the US actually built P51, P38 and P47 in parallel (not to mention F6F, F4U, etc). Unless the economics system makes that not a stupid option (assuming the US ordering system wasn't stupid), any game is just fantasy. Even limited shifts would worry me, although mostly because once you allow say +/-10% the debates on 'why no 11%, or 15%...' start.
Exactly. There are two separate (but related) problems. First, by virtue of hindsight, we know what the most successful designs will be, so it is easy to make all the "right" decisions in terms of production, way ahead of schedule. Second, and more game-specific, it is easy to max/min the attributes of each weapons system based on the game data and mechanics, so you can end up with an army stacked with a bizarre and historically impossible allocation of weapons -- like 33% Hetzers, 33% Flammpanzers and 33% 75mm-sporting halftracks, with the Landsers carrying 50% SMGs and 50% flamethrowers. (NOTE -- these are not based on the WITE values, but on a certain other game that I have spent way too much time playing over the years).