This is actually one of the more interesting and under-discussed pixel peeping topics. DxO's comparisons of sensors of different size appear valid and reproducible, unlike the immeasurable and vague terms like "tonality" and "microcontrast" which too often creep into discussions when comparing formats.
The differences were driven home for me when I was comparing my 450D against a potential upgrade to 7D on DxO. The comparison was expectedly favoring 7D, until I changed to the confusingly-titled "print" mode, which compares the sensors as if printed at 300 DPI. The differences in tonality, SnR and color sensitivity all but disappeared at the ISOs I use (100, sometimes up to 400).
The only advantage 7D has left with are a one stop more DR and half a "stop" higher megapixel count. While those are nothing to scoff at, they are not enough to justify upgrading for me, even with 7D's improvements in UI. And I'd be moving up in size and weight, which was one of the main reasons why I moved from 30D to 450D in the first place.
It seems as though Canon has managed to squeeze similar performance out of a much higher density sensor with their latest generation, which is no mean feat. But I would much prefer true improvements which can be seen in prints rather than only at 100% magnification.
Back to the OP's question, I don't see why similar performance would and could not translate to larger sensors, but I assume it has to do with cost. I'm pure layman when it comes to sensor manufacturing, but I imagine the cost of scaling the small high-performance sensors to FF would be prohibitive, as yields from larger sensor dies would presumably go down. Also, the manufacturers of some of these sensors are different, so the FF manufacturers will not have access to some of the features found in the smaller ones.
