http://new-astronomy-ccdcalc.software.informer.com/1.5/
This is a free calculator that helps work these things out.
Basically .5 to 1 arc sec/pixel is good practice.
Your 5.2 nm Canon is at .86 arc seconds/pixel.
The KAF8300 seems to work with most scopes so is a pretty safe bet and that is 4.54 microns. It also works well with short focal length scopes like FSQ106 etc.
Undersampling is common with short focal length scopes and as your article says there are plenty of good FSQ106ED and 9 micron CCDs like SBig STL11 or Proline 16803 images.
If you over sample you lose some sensitivity and takes longer exposure time to achieve the same signal to noise. Not much of a problem.
Small pixels have smaller full well capacities but with Sony low read noise chips stacking shorter exposures means there is no disadvantage there.
So in short your 10/12mp one shot colour APSc sized ccd you mentioned
should be fine.
One shot colour is not so good for Ha - no. Because only one in 4 of the pixels is registering red which is where Ha is. Mono is quite a bit superior for that. One shot colour is handy but from my experience it means worse signal in the dim areas which means more noise. So its fine for bright objects but if you start imaging things that are dim or Ha O111 narrowband imaging they aren't so good.
Starlight Express MX25C has a good reputation and high sensitivity. Also the Sony ICX694 or 814 colour should have good sensitivity as the mono version is one of the most sensitive ccds to narrowband.
Greg.