I added a few 67 lenses and create a separate Flickr account to just store the star test images. You can see them here:
https://www.flickr.com/photos/140916263@N04/sets/72157665841471914
Most of the test images are made from stacks of very short exposures (2 to 6 seconds). Later I realize that this cannot reveal color aberration, as shown by the P67 200mm/F4 case. So at some point all these tests need to be redone with longer exposures. This will be time consuming and also slightly more technically challenging. So I am not sure when I will have chances to do this, but I will try.
One issue I found long time ago and confirmed by these tests is that P67 lenses become out of collimation very easily. A good fraction of the tested lenses are out of collimation. The symptoms are not-round stars in the center or the shapes of stars in two opposite corners not mirroring each other. In the Flickr pages, you can easily see this on the color fringes of the 67 200/F4 and the shape of stars of the 67 75/2.8 AL. This is also true for my 67 ED 300/F4 on D800. So you should be really careful when you invest on a used 67 lens. The lenses may be good initially, but not so after a few years of usage.
Based on the data so far, and assuming that color aberration should be less an issue on shorter lenses, my ranking of the lenses tested so far for astrophotography is (from high to low):
1. 645 FA35/3.5 (needs to stop down to F5 or F5.6)
2. 645 DFA55/2.8 (needs to stop down to F5 or F5.6)
3. 67 90/2.8 (needs to stop down to F4.5)
4. 67 165/2.8 (needs to stop down to F4.5)
5. 67 75/2.8 AL (needs to stop down to F5.6)
6. 67 105/2.8 (needs to stop down to F5.6)
7. 67 200/4 (strong color fringing that doesn't go away at F5.6)
If I have to draw a line, I would say #6 and #7 are not capable of doing good astrophotography on 645z. #5 is the minimum I can accept, if the copy does not have collimation issues.
I hope at some point I can put my hand on the 645 DFA35, but this won't be any time soon, unfortunately.