(at least before you start cropping, which is of course a whole other subject)
But the cropping is the whole point of the
crop factor "argument". If your glass is too short then you will be cropping. Stick a 400mm lens on a 5D2, 1D3, D300, 50D or something from Olympus and if 400mm is too short on the Olympus it will be too short on all the others too. You will end up performing a software crop on images from all the cameras and you will end up with exactly the same useful area of sensor remaining, regardless of which camera you used. It might be 12x8mm; It might be 8x6mm, but to achieve the same composition you will end up with the same area of sensor being used from all the cameras. At that point the discussion about crop factors is moot. All the cameras will have been software cropped to exactly the same crop factor, meaning no difference whatsoever in terms of image size from any of them.
The idea that "crop factor" contributes to some sort of image magnification is preposterous. It doesn't. All that crop factor does is to narrow your angle of view, save you money on the camera and possibly reduce the file sizes you have to process. A crop viewfinder will make it more difficult to locate a small bird in the frame when using a long lens, possibly very difficult at times. A camera with a larger sensor, wider angle of view, and full 100% viewfinder display will assist you in locating your subject quickly.
As for "small" sensors having the worst IQ, again, if you are cropping images from each sensor down to the same physical size, in mm, then to all intents and purposes all your sensors are the same size. If the pixel structure was identical between cameras and the only difference was the sensor size then there wouldn't be a scrap of difference in IQ from any of them, once you'd cropped the image down to the same sensor area.
What then differentiates the cameras is the IQ per square mm of sensor area and that is a whole lot harder to get to grips with since there are so many variables to consider.
At the end of all that, let's not forget that there may be times when your lens is far from too short, and you have no trouble filling the frame. Well, now it's a very different story. The larger the sensor area you can project your subject/scene onto, the more light you will gather and the better the IQ you will achieve. This is the time when the limitations of little sensors reveal themselves and the larger sensor take an overwhelming lead in IQ.
IQ is principally driven by the amount of light gathered. You gather more light by projecting a larger image onto a larger sensor area. To get a larger image you need longer glass. Fiddling around worrying about pixels, pixel densities and per pixel noise is neither here nor there, relatively speaking. What counts, primarily, is the area of sensor put to good use. It's that simple.
Is a 4/3 camera with a 50mm lens going to give a 5D, 5D2, 1Ds3, D3 or D3x and 100mm lens a run for its money? No, not in a million years. A 4/3 sensor has approx 40% of the area of an APS-C sensor, which itself has an area only ~40% as large as a full frame camera. 40% of 40% is only 16%, a lot better than a point and shoot but a long way from the big boys.
Buying into 4/3 is likely retreating away from full frame and heading back into the woods with the point and shoot cameras. It's the wrong direction to be heading in if IQ is your concern. If cropping tighter was the solution to improving IQ I'm not sure there would be the market there is for 500/4, 600/4, 800/5.6 lenses and teleconverters on top. Where is the upgrade path from the 4/3 system?
Now, I don't want to be mean about the 4/3 system, but its appeal must surely lie in its compact size and modest entry level price for a body and quite a bit of focal length. But where do you go when 300mm is no longer enough? Remember, the magnification from a 300mm lens on an Olympus body is no different than 300mm on a Canon or Nikon, despite the impression the "crop factor" gives to some people.