I also just read that article, and it seemed to me that the author neglected to attempt to try his microfinished HS's WITHOUT thermal compound when attaching them to his crystals. He seemed to be implying that the thermal compound interfered with heat transfer when the HS's were "too" smooth, because it couldn't be squeezed out of the way when there weren't enough peaks and valleys to hold the excess. This would imply that the TC would not be necessary if the mating surfaces were so smooth that the TC got in the way. We all know that TC is only needed to fill the voids of imperfect finishing, and that even the best TC is not as good as direct HS to substrate contact. Otherwise, we'd all be using gobs of TC between our HS's and CPU's. This is a situation where sometimes "less is more", and by extension, it would seem possible to reach a point where "none is the most". Of course, I'm not advocating that everyone throw their Arctic Silver in the nearest trash bin, but it would appear (based on this article) that we could reach a point of diminishing to negative returns by the use of any TC.
I realize that the author apparently had access to very sophisticated machinery to flatten and polish his HS's, which most of us would not, but it raises the possibility that if the HS and CPU are both highly polished and flat, that the TC might become an unnecessary component in the equation.
Perhaps an experiment could be done to see if once a certain smoothness/flatness spec is reached, that TC becomes a hindrance, rather than a help in facillatating heat transfer.
Just a little food for thought.