If you go to the bottom of the article, those sensor sizes are addressed. For the 1/1.8" you note with 10mp, they show the optimum is only 4mp! for the 1/2.5 with 8, should have no more than 2.7mp. So, yes they do address your point.
More assumptions on their part they are assuming that the chip making technology of ten years ago will stay the same.
The 1/ 2.5 sensor should be limited to 2.7mp is not a standard, it is technology driven.
The articles optimum is not the standard, they would have put more pixels on the sensor if they could. Back then the individual size of a transistor was much larger than it is today.
2.7 mp was the most amount of pixels they can place on a 1/ 2.5 sensor ten years ago, today 8.0 mp can be placed on the same sensor and looks to be the limit without going over the ISO standard noise limitations.
Anything over 8.0 mp over the sensor size will have to grow, just as Canon had done.
The article also assumes that the individual cell sites or pixels on the sensor will never improve in its efficiency. Meaning that that ability of each pixel to absorb or detect the incoming light information will stay the same.
Keep in mind that the article is addressing the noise issue and not the resolution. The resolution between a 6 mp and 12 mp has a huge difference when we start blowing things up.
Good examples are the Nikon D70 at 6 or 7 mp and the latest D300 at 12 mp, the sensors physical size are the same, the D300 has twice as many pixels but they are smaller in size I dont see too many folks buying up the D70 because of its better picture quality over the D300.
Sam