Having worked for Texas Instruments for IC and Darlington production I doubt that an efficient masking technology could be developed for spherical surfaces, especially for today's requirements of generally less than 1-micron resolution. Etching could be another substantial issue as depth is to be measured in reference to a plane.
Also, such a curved surface would work fine with lenses of over 45-50 degrees, but what about the teles?
I think this was one of the reasons why the major players were (apparently) cautious in jumping into the mirrorless sector; what the Nex-7 wrt the Nex-5N has pointed to: Offset microlenses seem to be the only solution and not a costly one; however they function well with CCD-cells but not so well with the CMOS.. Certainly there should be a R&D going on in regard to this issue.
Also, pure conventional designs might not work well on the CCD-microlenses too (Biogon 21/4.5 or Summilux 24/1.4 corner smearing or color shift issues.. ) The best design approach is from Fujifilm XP1, totally custom design with huge rear elements however they were to cover only APS-C format.
IMHO, there could be also another factor for this: The future full-size due to AF and size requirements could well be the APS-C size.. Today the APS-C can easily satisfy the ISO-6400 capability, resolution of 24MP (next year most certainly over 30MP) and the lens design for this format could be less problematic than the one for FF.. I think Leica too is working on this format for their future models, we might even see a concept camera in this Photokina. (AF is unavoidable for any company to survive.)
