My understanding is that the read noise in almost all modern CCD and CMOS image sensors is about as low as it's going to get but there may still be a little improvement to be had (e.g. the 1D4 sensor and the latest Sony sensors in Nikon D7000 and Sony A77) so we'll almost certainly see this in the all the next Canon sensors.
Plenty of room for improvement between the 7D and D7000, especially at low ISO.
What do you mean by "plenty"? We're talking about read noise at low ISO right? Look at the info you linked to... 7D read noise is 8.6 electrons and the D7000 is 3.1 electrons. So the difference is 5.5 electrons. So there is a little room for improvement which as I said we've now seen in the latest Sony sensors. By the math, this is a a little over 1 stop of DR.
However, the increase in DR due to the lower read noise is being offset by increasing resolution (smaller pixels) so overall we might see at most another stop of DR in the next 7D but maybe not.
No, this is a myth (hard as hell to kill, too). Smaller pixels don't make poorer DR inherently.
So, 7D and G12 both have 11.2 stops of DR at base ISO, but the G12's pixels are 4.4 times smaller.
Hard to kill because it is NOT a myth. First let me point out that the G12 sensor is a CCD and the 7D sensor is CMOS so not apples to apples. CCD are much simpler and the photosites can be made larger for any given resolution (more specifically "pixel pitch" or pixels per unit area on the sensor). This is one of the big advantages of CCD. The other advantage of CCD is that they have lower read noise (actually did have
but CMOS has caught up in the last few years which you see in the Sony sensor data in your link) . The combination of larger photosites and lower read noise gives much higher DR than CMOS. When the industry first started to move to CMOS it was because of expected lower costs but that turned out not to be true at first but did come down over time (CCD has come down in cost as well but not as much). CMOS has now caught up to CCD in read noise but not in photosite size. The photosites are smaller with CMOS because there are more electronics required on the sensor itself and in fact take up about half the surface of the sensor. CCDs don't have as much electronics on the sensor and the photosites can be made larger for a given resolution.
DR is better with larger photosites. The larger the photosite the more electrons (converted/released from absorbed photons) can be collected so that the saturation level (the largest number of electrons that bucket can hold) is higher. Look at the links you provided... the saturation levels are higher in the 7D and even higher in the D7000. The D7000 has the highest saturation level and low read noise and therefore has the highest DR.
We can also see that the Sony sensor has much higher saturation levels than the 7D with about the same resolution and sensor size. So we can conclude that the latest Sony designs have increased the size of the photosites (i.e. each photosite can hold more electrons). I'm not aware of any proprietary/patented breakthrough from Sony so I think it's just the next evolution in the CMOS designs and we'll see it in the next Canon sensors as well but I'm guessing.
A good example of the progress in CCD sensors is to look at the data for the G10 compared to G11. Canon wanted to make a the sensor better so they reduced the pixel pitch but also increased the photosite size which is evident from the large increase in saturation levels. Also they got the noise down and the result is going from 9.3 stops DR in the G10 to 11.2 stops in the G11. I'd like to see the sensor data for the new CMOS sensor in the S100.
f/8 autofocus is possible but what's left to differentiate the 7D2 from the 1D4/5?
The 1-series has a built in portrait grip, a bigger battery, is built like a tank, has better weather sealing, has UI designed to be used with gloves on, has a larger sensor, produces faster frame rates, and has longer shutter life.
I meant my answer as a bit of joke but it didn't come out that way I guess. Of course there is a lot to differentiate the 1D4. And with the latest processing speeds and new sensors the 1D5 will probably be FF because they will be able to get 10 fps from a high resolution sensor. I do agree that if the for any sports oriented camera (1D4, 7D, and successors) AF performance is key and will always be an area for improvement.