The main reason for retrofocus design in RF wideangle lenses is that it offers better control of aberrations. The symmetrical design had approached the theoretical limts of correction, one of the reasons the Elmarit 21 was not renewed for so many years. I am sure that Zeiss and Leica could design a 2.0 21 within the size-limits imposed by the body design of their rangefinders on paper, however I doubt that the tolerances of the mechanical part of the lenses, which is magnitidudes more critical in rangefinder lenses than in SLR lenses for various reasons,plus the shaping of the glass could be realised for an acceptable price, even by Leica/Zeiss standards.
Ironically, the design of a high-speed 21 on a 35 mm sensor could be far more difficult than a 16 or 18 of the same speed on a 1.33 sensor, due to edge and corner problems, that increase exponentially with both sensor size and lens speed. As a comparison: The Summarit was designed as a 1.5 because that was the limit of possibilities at the time, and 1.4 was totally impossible. Such a small difference!
On a more optimistic note: If we consider the advances in lens design and manufacture over the last, say, 15 years which are not far short of miraculous, especially with Leica, Zeiss and to a certain extent CV, it may well be that high-speed short lenses will appear in the not too far future. Certainly the advent of digital, smaller sensor rangefinders will speed up this development.
