Aero downforce versus drag is a pretty fine line to walk without a huge amount of testing and development that even the most well financed race teams in the world struggle with.
Aero tricks can certainly work at almost any speed if they are designed with that in mind; you do not need to be going very fast to generate some effect. Having the car stuck down with almost always help in any situation more than it would hurt, overall. I think that the effects of a fastback added to a roadster would be less noticeable than adding radical downforce generating elements, since the rules are wide open...
Drag on top speed and acceleration can be considerable and felt at low speeds and is mostly a function of total frontal area. Cutting out a large windshield will lower that area some but will hurt the CD; I couldn't even really quantify the effects without some expensive software and and engineer to run it. If you are not spending a lot of time total near top speed than I would worry less about the size and shape of the hole you are punching in the air. Here, weight is also a factor and less weight wins a whole lot of battles...
My big though: I have seen events where cars run a road course one day and a standing mile the next. Smart racers have adjustable aero elements and can add/remove them easily. I think this is where you need to think some. Chopping a windshield is a big move, same for the the work and expense of adding a fastback. Neither of them alone is going to override the effects you can otherwise generate with an unlimited rule class. That being said; I always tend to lean in the direction of less weight and especially if high up off the ground.
I know that isn't an answer or even clear opinion, but maybe it isn't the proper question for your usage? 
