Joe -
here are some Thermocouples you might find interesting that will fit beside the die as I illustrated a ways back
Phaestus - well written, interesting question regarding the Duron. I like your writing style
One thing regarding consistency - will you not have most of the same issues regarding replicating the setup on a die simulator as you do on a real CPU? (thickness of thermal compound, mounting of the block, etc.)
Regarding the variability of the diodes from part to part - I can give you a quantitative answer to that in degrees C if you are interested or provide you with the info on how to determine it if you prefer. It might well be within the measurement error of the setup.
Question for you, honestly interested in your view. Which problem(s) do
you think is(are) insurmountable?