Michael, allow me the indulgence of posting an article I had prepared for another forum.
1. You already have a lens system which is optimized for 35mm. Every glass in your range (except (except for newly introduced glass, like the EF-S and DX) delivers an image circle for a 35mm frame. I would argue that EF-S and DX is a kludge, definition of a kludge being "an error introduced to fix another error". If Hasselblad wanted to produce a digital back for its MF camera, would it buy a 6.0MP APS sized sensor from Sony? The answer is "no" for many reasons, but one of the prime reasons is because too much of the imaging circle has been wasted. The argument remains true, in a smaller scale, for FF vs. APS sized sensors.
2. Expanding on my first point, the greatest expense in designing a lens is to ensure the lens performs consistently from corner to corner at wide apertures. This is one of the touted advantages of DX/EF-S/Four-thirds lenses. Paying a small fortune for a 600 F/4 and then throwing away the most expensive part of the image sounds wasteful to me. This would not be so bad if the manufacturer rationalised the entire lens range so that savings could be realized across all the lenses - something that the 4/3 consortium is trying to achieve, something Nikon is halfheartedly doing, and something that Canon will not do.
3. To continue along this line, where are the savings which reduced crop lenses have promised us? First 3 examples, price obtained from B&H, last example, Canon Australia website:
Nikon 10.5mm DX fisheye, USD599.95 - Nikon 16mm fisheye (full frame) USD$564.95
Nikon 17-55/2.8 DX, USD$1349.95 - Nikon 28-70/2.8 (full frame) USD$1329.95
Canon EF-S 10-22/3.5-4.5 USD$599 - Canon 17-40/4L (full frame) USD$674.95 (not strictly comparable, since the 17-40/4L is a constant aperture lens)
Canon EF-S 17-85/4.5-5.6 IS AUD$1099.00 - Canon 28-135/3.5-5.6 IS AUD$999
... I am sure you will agree there is no difference. We are still paying the same price for reduced crop lenses as we were for full frame lenses.
4. The next part of my argument concerns sensor design. Among other things, sensor designers are concerned about resolution, dynamic range, and noise suppression. Pixel pitch (the size of each individual photo pixel sensor, in micrometers) is just one out of many factors which influence these. People like to use the rainwater and bucket analogy which is a good one - a larger bucket allows more precise measurement of rainfall. A small bucket spills water too quickly. If there is very little rainfall, the small bucket may not gather any rainwater at all.
You can design a more efficient water gathering system (microlenses), you can design a better ruler (both tricks employed in the new 20D and 1DMk2) ... but in the end bigger buckets are still better.
A full frame sensor ultimately allows you to fit more, and larger pixels. Eventually you will hit a limit on a smaller sensor before the conflicting demands of resolution and larger sensors force you to look at other means of increasing dynamic range and reducing noise. You just hit the limit later on a FF sensor. If two manufacturers have exactly the same noise reduction and image processing technology, the manufacturer that can produce a larger sensor will have higher resolution, increased dynamic range, and less noise.
5. Viewfinders. Its not strictly necessary to have a larger viewfinder - it's just nice to have.