What would happen if a manufacturer using 16x9 native panels decided to use an anamorphic lens (let's say it expanded horizontally) as the spec lens. Not a second lens, the prime lens. Then used some software that could feed what we usually term 'squeezed' frames to the lens to send 2.35 to a CIH screen. Then, when the native material was 16x9, a graphics processor could output a correctly proportioned image on the same 2.35 CIH screen. Yes, you would have unused bars to the sides.
The prime anamorphic lens would be a little bigger than the spherical prinme lens would be, but not by a great deal.
The anamorphic elements would be placed early in the chain, mostly near to the back of the lens and thus wouldn't need to be big at all. This is done routinely with camera anamorphics and commercial cinema anamorphic projection lenses and works fairly well.
The elements after the anamorphic elements perform some correction for aberrations arising from the anamorphic elements, especially geometric ones like pincushion. There is a pincushion improvement but not necessarily elimination. Usually there is a sweet spot - an ideal throw and throw ratio - where everything comes together, but outside that range you see some fall-off which, is why companies like Schneider make a range of these types of anamorphic primes.
The following link contains a list of 22 such lenses made by Schneiderhttp://www.schneiderkreuznach.com/pr...anamorphic.htm
Schneider patents describe a lens system where the spherical component of the lens deliberately aberrates the beam in one "direction" so that unavoidable aberration by the anamorphic component in the other "direction" is balanced. If either section is used on its own the picture looks terrible, but together they work to produce a clean image.