The effect of the backside illumination is rather small, when a large scale sensor is used. It could be, that backlit-CMOS technology produces a measurable higher sensitivity, but never ever will this be doubled compared to conventional technology.
"Back lighting of the sensor improves peripheral performance with ultrawide lenses"
How should this work?
First of all, back illumination is not new - it has been standard technology on high end CCD (not CMOS) sensors for at least 10 or 15 years. Moving the technology over to CMOS is a logical thing to do and is the kind of thing that could be expected in a very high end camera. That said, large, back illuminated CCD's are horrendously expensive - it's moot whether they would fall within the typical pro budget.
Secondly, the effect on sensitivity of back illumination is dependent on the ratio of capture area to gate area, so in the era of microlenses it is not particularly signficant. (Judging again from CCD's you get 20% to 30% at most.) 30% would be about 0.4 stops which is not to be sneezed at, but isn't going to save the planet either.
Thirdly, it directly affects your dynamic range because the volume of the charge trap is intrinsically larger than the front illuminated chip. This matters a lot - it might provide a true 14 bits.
Fourth, it _could_ improve performance with wide angle lenses because light striking the lens at a large angle is less likely to be shadowed by the gate structure and thus more likely to be captured by the sensor. I suspect this would be a small effect. Having the sensor near the surface would definitely allow for better microlens design so that light ended up where it was supposed to.
So - my version of the post...
Is back-illumination possible? I think it is.
Is back-illumination likely? I would guess that it's already been done but it's not remotely affordable.
Is it the correct solution? No, I think a detector cascade aka Foveon) is a far, far better approach.