I was somewhat "forced into upgrade" because I had the shutter fail on my K10D the week the K7 became available in stores. Odd coincidence isn';t it.
I then determined that the repair price for the K10D was not too bad $325 for shutter mirror assembly and lens mount plus total cleaning and overhaul that I sent it in after picking up my K7.
I chose the K7 over K20 for a variety of reasons, principally the improved metering.
The K10D has the worst metering using legacy lenses with manual apertures (of which I have 20), and when using teleconverters that report aperture direct from the lens without modification. (I use sigma 1.4x and 2x TCs on a sigma 70-200F2.8). the K20D has all the same problems.
Early test reports from forum members showed the K7 metered significantly better, and my own tests since have confirmed this point. metering is greatly improved for manual aperture lenses below F5.6 which also resolves the issue of using teleconverters.
The K7 also has other features which are greatly improved for someone who uses jpegs right out of the camera, including extended dynamic range with highlight and shadow detail protection (when used) and 3 levels of sharpness control.
The other thing to consider is the 77 element metering matrix, (up from 11 on the previous cameras)
All else aside, I don't really think there is much difference between the K20 and K7 in terms of resolution and high ISO performance, after all they both are 14MP CMOS sensors, so that to me is anon issue, although in principle, the K7 should be a little better for noise because it is intended to support video which leaves the sensor powered longer, hence in principle one may expect thermal management to be better.
Having said all that, at this point, it might be interesting to wait, and see what comes next. Except for thelag between *istD and K10D pentax has updated their top end camera annually so we are overdue a little.