This kind of thing already happens in established fields of forensics all time and if they couch everything in enough authoritative sounding language nobody who doesn’t already know a whole bunch about the topic is going to be able to call them out on it.

Bingo, the goal is not to actually trace 3d printed guns, it's to add another piece of evidence they can slap onto the target of their investigation to improve the chance of getting a conviction. Whether the accused is actually guilty of anything is truly irrelevant to the intended purpose of these kind of things, they are in practice assumed guilty by the prosecution unless proven innocent. The point is simply to make a convincing case to a judge or jury.

Just change nozzles every print, you can get cheap Chinese nozles for less than a $1 a pop.

I think there was a study that showed every 3d printer leaves a sort of fingerprint, and they can identify which printer printed what object.

You don't have to buy it for them to convince 12 average citizens that they used wizardry and science to definitively prove it's 'real.' See bite mark analysis, fiber analysis, hair analysis, etc, etc, etc. Court rooms are full of pseudoscience because it helps convict more people.

Hell, you're not even gonna get that far. I don't think there's a forensic lab within 2000 miles of me that could do this kind of work. All that funding goes to guns and cover ups.

You could probably map resonance artifacts, but you have to isolate layers that were printed at the same speed and direction. However, the second you tighten a belt or screw, that pattern will change and I am not sure how consistent resonance patterns would be on a bed slinger. (The quantity and density of printed plastic may change the resonant characteristics of the entire printer. This may be less of an issue on a core xy.)

Thinking waaay outside the box.. In some cases, I have seen extruder gear marks on the filament create artifacts on a print. Every gear pattern should be unique, but measurable differences would probably be micron or sub-micron.

Maybe you could map the surface of textured beds as I seriously doubt that those patterns would be consistent and more prone to randomness from the factory.

There are a ton of conditions that could generate unique artifacts on a print, now that I think of it. Hell, even a printers PID tuning can leave visible and repeatable errors.

I half expect Bambu labs to already do and be furiously trying to come up with a scheme to monetize it.