That X-ray technique is very cool. One problem, though, is that it only shows the metal layers. The doping of the silicon is very important, but doesn't show up on X-rays.
I haven't ever looked into it itself, but what you just pasted seems like a somewhat promising answer indeed. Except for the synchrotron part I guess? Maybe?
They were using 6.2keV x-rays from the Swiss Light Source, a multi-billion Euro scientific installation. 6.2keV isn't especially energetic by x-ray standards, (a tungsten target x-ray tube will do ten times that) so either they needed monochromacy or the high flux you can only get from a building-sized synchrotron. Given that the paper says they poured 76 million Grays of ionizing radiation into an area of 90,000 cubic micrometers over the course of 60 hours suggests the latter. (A fatal dose of whole-body radiation to a human is about 5 Grays. This is not a tomography technique that will ever be applied to a living subject, though there is some interesting things no doubt being done right now to frozen bacteria or viruses.)
"Though the group tested the technique on a chip made using a 16-nanometer process technology, it will be able to comfortably handle those made using the new 7-nanometer process technology, where the minimum distance between metal lines is around 35 to 40 nanometers."
The discernible feature size required for imaging CPUs of 6502 or even 8086 vintage is much lower than that. The latter was apparently made with a 3 µm process, so barring any differences in process naming since, that's about 3 orders of magnitude less.
Plus, as Ken's article says, the 8086 has only one metal layer instead of a dozen.
So my (still ignorant) hope is that you maybe don't need a synchrotron for that.
I don't know about a hobbyist, but are you talking about something along the lines of "Ptychographic X-ray Laminography"? [0] [1]
[0] https://spectrum.ieee.org/xray-tech-lays-chip-secrets-bare
[1] https://www.nature.com/articles/s41928-019-0309-z