but the OS only sees the LBA Flash transition layer which converts the page sizes to the OS sectors so it should still work
yep.. that would be up to the SSD controller to spread the logical to where it saw fit.
If I had to guess why we probably won't see this type of implementation?.. increased logistics, added firmware development, added hardware complexity, to ultimately mean higher cost for insufficient gain/return.
And it's actually quite a good idea as the mfgrs could better tune the drive to function at the most efficeint level of performance based on the particular data sets/load. I've seen and tested page size changes and am familiar enough with them to have seen some distinct trends in performance variation(you gain in some metrics but lose in others). In that respect they appear to be very much like multi-displacement engines in cars with the ability to become noticably more efficient in how they deal with engine loads.
4 core SSD controllers or at least higher clock speeds on the ones we have available so far, ONFI 3.0 spec'd nand, variable page size capability, larger DRAM, and kick-ass firware designed to take advantage of it all. Never hurts to dream, right?