This comparison with EMCCD, in my opinion, is biased. A large pixel can always beat a small pixel sensor given a equivalent QE and read noise. All single photon detection these days are using EMCCD or sCMOS. Canon should publish its QE curve, read noise and dark current data.
Actually if you do the modification yourself by just removing inner IR absorption glass (Color correction Filter) while preserving the Dust reduction filter, you will have at least 4~5 fold gain in sensitivity at H-Alpha. The ICF in 20Da has only 80% transmittance instead of 99% for an dedicated astrophotographical modification.
Kodak may not be a success in the commercial market, yet it provided numerous solutions for scientific instrument. If you know astronomy, almost all affordable CCD cameras for amateurs and professionals use Kodak chip. Apogee, FLI, SBIG, some models from QHY and ATIK all use kodak CCD. Even the Illumina Sequencing Analyzer in my lab use the kodak CCD. Though inferior in terms of read noise, and dark current compared to Sony, they provide all kinds of Black and White sensors in various format and pixel density. Sony only put its newest technology into high density sensor for cellphone and Point and shoot, instead of large pixel low light sensor. With the diffraction limited in place, the 24MP A77 is just another bad example.