Have you looked at those conic section inside a slightly larger conic section void type grinders? Used often for grinding the shells off macadamia nuts, because smaller parts fall through and larger parts are cracked by the inertia of the rotating inner section. 

The coming video explains why coffee grinders are for coffee beans, flour mills for flour etcetera and why it is so hard to grind plastics at room temperature. Stay tuned!

The new blog entry on this project gives hopefully a better idea of the energy needed to grind PLA pellets into plastic powder:
https://hackaday.io/project/181165-my-machinery-2021-direct-granules-extruder/log/228264-lets-have-a-look-at-the-energy-needed-to-grind-pla-pellets

Yes... well... very good video but none of those are at all related to the conic section grinders I was suggesting, and coffee beans or wheat grains are not comparable to macadamia nuts when they still have their shells on. They are by far the hardest nuts to shell needing over 200,000 kg / m^2 to crack and required the invention of a new type of grinder to both provide the energy required to crack the shell, and still not get bogged down when processing many nuts at once. The point of this type of grinder is that it applies NO force to the smaller particles and applies the entire force of the rotating mass to the larger particles until they have cracked. It is a very selective system, not at all random. The cracked nuts quickly pass out of the system at the bottom because they fall through; they do not depend on being blown out the side, so they do not contribute to drag once they have been split. This is critical for the nuts, because the goal is to get the meat of the nut out whole, having only removed the shell. Applied to grinding PLA, it should result in a very consistent size of finished product.More than that, because the conic sections are round and blunt, they require no sharpening or inherently delicate geometry. All that is needed is a slightly rough surface to prevent the shell from sliding which can increase heat. Cast cement or iron or some combination, e.g. inner cone of cast iron, outer of cement. My objection to the idea would be that it's somewhat difficult to fabricate... but I think a cast cement might work without putting too much garbage into the result. You are, of course, the person doing the work, so it is for you to decide what is a pink unicorn and what is not, but I see no evidence which would challenge my small suggestion. 

Okay, those macadamia nut crackers seem to be very special. Do you have a link to such a machine?
But concrete is for sure no material for a PLA grinder. You will end up with concrete in your plastic powder which will cause clogging of the printer nozzle.

The kind I've seen is shown on page 65 but labeled as a "husker" while I'm sure I've seen that sort used as a cracker, but without the spring loaded sides. The cracker shown on 67 is another possibility; a set of horizontal rollers that don't quite touch. It's made metal cylinder with a cement core so no cement contacts the product. 
https://www.ctahr.hawaii.edu/oc/freepubs/pdf/res-039.pdf
Another of the horizontal ones is patented here:
https://patents.google.com/patent/US2230790A/en
It looks like most of the modern versions are this new horizontal, dual roller, type, which is different from what I remember, but has the same idea: A heavy rotating cylinder which provides an "unstoppable force" and some other very strong object which provides the "immovable object" which the product is crushed against. In this case, both rollers fill the role of the  unstoppable force as well as the immovable object.

Thanks for pointing to the documents!
These machines are indeed totally different from any shredding / grinding / milling machine I knew of. Not really simple mechanics.

Very interesting analysis of the problem and solutions!

Thanks! More on the general problems of grinding PLA pellets will follow soon.