After one or two bits of infrastructure-building - figuring out the format of the sensor-data part of Nikon's .NEF files (the last 9728000 bytes of the file contain 2000 lines, each of 3040 samples packed as 304 16-byte chunks, with each chunk containing ten 12-bit numbers packed two-in-three-bytes as 0xAA 0xAB 0xBB and a final 0x00) and implementing dark-field subtraction because the D100's sensor, especially at high ISO, is plagued with hot pixels, I got some reasonable shots.
The problem is, the star images are big and ugly. I'd calculated, based on a pixel size of 7.8 microns, that a star would move three pixels in a 1.6-second exposure at 170mm, or one pixel in a 0.3-second exposure at 500mm. I'd noticed that it took a while for vibrations in the tripod to settle, so used self-timer to let it settle before taking the shot. But still a star at 500mm is fifteen pixels across on the 0.3s-exposed image. Seeing in Cambridge isn't great, but it's not likely to be as dreadful as 45 arc-seconds.
My guess is that I'm not focussing correctly. But I'm not sure how I focus correctly; there's no light, so auto-focus just hunts back and forth and gets nowhere, turning the focus dial until it hits the infinity end-stop doesn't seem to be sufficient; my eyes can't distinguish a perfectly-focussed star from a slightly out-of-focus star. Towards the end of the session, taking photos pointed pretty much straight up with the lens at 500mm, I was getting images trailed by movement of the tripod rather than the stars, but that's more a matter of waiting for a season where the interesting object isn't at the zenith.
[ For taking pictures of stars, I think the figure-of-merit is focal length / f_stop^2; the permitted exposure before the motion of the stars blurs them is proportional to 1/focal_length, and the amount of light that gets in is proportional to the lens area = (fl/ap)^2. For star-fields, it's 1/(fstop^2 * focal length), since the area of sky you see, and so the number of stars, is proportional to 1/fl^2. 50/1.4 lenses are what this measure tells one to lust after ]