This is what Tesla tries to do, though. They produce a vanity project for a high price and proof-of-concept and then they leverage that experience to produce more economically feasible versions. Once the Model 3 comes out we will have some idea if the concept can work.
As far as roof penetrations go, I assumed that each tile would lock into another creating the circuit in the tiles themselves, minimizing the need for roof penetrations, and that tiles would then be connected to sensors, or the circuit would have diagnostic software, that could tell where in that circuit had problems. Is that not the way that they work? I guess we don't know yet.
