It seems like every time that someone announces a non-Bayer sensor the Internet gets all abuzz with hyperbole and overfed expectations. The latest Sony story got picked up on 14 RSS feeds I monitor (and rising).
Sony's announcement was about an RGBW sensor. The W is "no filter." Besides the fact that it's a bit of a replay of the Kodak Monochrome arrangement announced a couple of years back, you don't abandon Bayer without having other implications. While a non-filtered photosite does give you more luminance data, that only helps when it doesn't clip ;~). In bright situations, it's likely to clip. Indeed, if you set your W exposure so it doesn't clip, then you have trouble with color, where there will be more noise (chroma noise, the worst kind).
Others were excited because they thought Sony's announcement would impact serious cameras. Not likely. What most didn't notice (because they got summaries of English translations) is that this new sensor is attempting to give 1.4 micron photosite performance in a 1.1 micron photosite. In other words, camera phones and compact camera sensors with more pixels, but "same" image quality. Only problem is, it won't be same image quality for reasons noted above.
Sensor progress is relatively constant. We get small gains in efficiency, read noise, light gates, photosite isolation, well capacity, microlenses, or any of the other elements that already exist in our sensors as we've got a huge number of organizations iterating those things.
To get a big bump instead of the small steps forward we're used to, I think it's going to take one of the disruptive technologies lurking in the background: quantum dots, full color photosites, light field microlenses, or perhaps moving from silicon to another more expensive and harder to work with material. Don't worry, you'll know it when it happens, as we'll likely make a bigger than D3 step forward when it occurs.
Let me flip that to another context more relevant to this site: what if the D4 is simply equivalent to the D3s sensor with 15% more resolution? In other words, very equivalent results, but a marginal move forward in number of pixels. That's the level of change we typically see in sensors in a two-to-four year stretch. Not a bad change at all, but not the over hyped type of change that gets stirred up every time we have a "new sensor design" announced on the Web.