"What's the grain diameter? If it's larger than the depth of the
cracks and crevasses in the aluminum extruded heat sink, then you're
taking a step backwards."

Now you're getting into the RMS finish left by the machines. In mass production I doubt it is very good. It is likely the grain is fine enough, in fact I have seen the results of machining on the surface and it looks like manufacturers are derating enough or just don't care.

"The bulk of the heat is passed by metal to
metal contact. "

Which is why I apply as much pressure as possible before tightening any screws. The strip, some devices are mounted by clips and the pressure depends on its modulus of elasticity. One advantage is that over time it presses continually and closes the gap, if any.

There is this pink **** out there that has such a high viscosity that I consider it unacceptable. It is hard to squeeze out of the damn tube, as such it would take so much force to actually get the high spots down to metal to metal contact that it would probably damage the device.

I have used tools to compress these, and I know when to stop. When you tighten the vise grips or whatever a little bit more and get no more goop then it is pretty much done. In the case of a large STK IC for example, I will take the old part and put it on top of the new part so it spreads the force and isn't likely to break the case. Squeeze the whole shebang until no more comes out.

As far as the compound and its additives, diamond dust is probably the best.. Copper beats aluminum or zinc oxide, but diamond blows their doors off. It also has a very good dielectric strength except for blue diamond. Apparently whatever "impurity" is in there causes it. I don' t feel like looking up what that is, feldspar or some **** ? Doesn't matter. Diamond does have a cost though, just how much is that semiconductor worth ?

"The lack of flatness also causes problems when one tries to using such a heat sink as a lapping plate. "

I wouldn't do that with a computer CPU. I have done it a few times with other devices and that was only so mush, I did not really remove any significant material. Truth is I was low on the "bird****" so I thinned it out a bit.. Beats nothing and the guy didn't want to wait.

Actually I have cleaned a bunch of them with coffee filters. They do remove some metal. In fact the did it to VCR heads as well, seeing that black on it was not dirt, it was aluminum. It is time to stop, but the heads were about as clean as they'll ever get, and it may have enhanced head to tape contact a bit. Just don't take off too much or there will be a bunch of wear and there goes your gap and azimuth offset sooner.

"If you want to see how bad it can get, take a black felt tip pen and "paint" the top of the CPU or the heatsink black. Find a flat surface (glass is good). Drop a sheet of fine emery cloth (2000 grit) on the flat plate. "

Actually the coffee filter should work.

That sounds like scraping. you know you have machines, and the ways are straight. Well those are cut on machine with straight ways so they are straight. And the ways for that machine are cut on a... this could go on forever. What is the FIRST reference ?

Scraped on plates. People actually scrape them by hand. they have a special surface on which the high spots are perfectly flat to withing a millionth of an inch, but there are valleys for oil. A roller will roll perfectly straight, or any guide that does not damage the plate. How to achieve this ?

They take two plates that are somewhat flat and blue them and put them together, where the bluing is not is a high spot so they take that down with the scraping tool. The two plates eventually seem flat, but they might not be.. One could be perfectly concave and the other convex to the exact same degree so they mate, but are not straight. So they have to use a third plate. you can't have a concave and a convex plate math another plate, it is simply impossible. And with hand tools they get them within millionths of an inch. I have watched them, and watched them sell them. Most companies do not need that accuracy but some do. they cannot tolerate a copy of a copy of a copy of a copy of flatness, they want the original. My friend makes the originals.

This is more accurate than polishing granite plates or even countertops. usually those are done with optical flats and/or light at an oblique angle, along with a good eye. I got a few sample from my late friend who did those granite and marble tops at his day job, other times he was a sculptor. I commissioned him to do one but he died so it is not done. i use those samples to process food, meat usually because we buy in bulk. not easy lugging a piece into the kitchen but it is worth the trouble, and I like to use on the occasions when I make a real pizza. All the pizza places around here just get worse and worse. They mostly suck or charge an arm and a leg.

"thermal cycling tends to pump the grease away from hot spots
leaving rather voids"

Hopefully those voids are where metal to metal contact was actually achieved. They are usually under spring pressure.

"Thermal expansion in an extrusion isn't uniform and
you will see bending in the heatsink. "

Mass production. I am surprised things work as well as they do. but all the mid fi audio amps are junk. Of course they want things to last two days past the warranty. One day is pushing it a bit too much.