I entirely understand your question, ratpack, but unfortunately I do not know what design aspects would be responsible for the sort of differences between units that people (myself included) so often report. My first instinct upon encountering such noticeable differences is to check for mistakes in setting up the equipment - bass management issues (such as main speakers set to small and a sub set to on in a system without a sub), tone control settings that are not zeroed out, or some sort of odd EQ being introduced, for examples - since those are the most logical source of noticeable variances in sound. In the case of my experience with my co-worker's 1068, I had gone through those items in some depth, so I have a hard time attributing my experience in that case solely to a setting oversight.

Others with some hands-on audio electrical circuit design experience may be able to offer some hints. The only aspects that I have heard and can make a clear and straightforward case for would be noise floor - as an example, I'm far from the only person to comment on how absolultely dead quiet the 990 can be (when ground loops are cleared out of the system, of course), whereas the 950 had the teething pains of the red dot and blue dot revisions in an effort to eliminate a noise floor-related hiss. That doesn't really seem to relate well in my mind to issues such as one unit sounding "harsher" than another or any of the other sonic-related critiques that we so often hear about.