Incremental deposition may not yield a working 3D printer, but couldn't a small ensemble of machines construct all their parts? (Minus the chips, for now)
It's important to remember that the majority of the core technology in 3D printing today actually dates back to the late 1980s. We're starting to see some interesting developments in materials and capabilities, but there are still plenty of limitations that need to be overcome.
Spindles, motors, circuit boards with components on them and bearings, slides and so on are all multi material or very complex processes usually only doable if you produce a lot of something in one go.
Just try to think about what it would take to print something as trivial as lacquered copper wire for stepper motor windings or a circuit board with a reasonable level of integration.
And the biggest issue with that prediction is that there is no gain from it: printing the non-commodity parts is the whole trick to efficient 3D printing, mass produced parts will have incredible accuracy and very low pricing so use them when you can and 3D print the remainder.
To me the biggest hurdle is the electronics, which currently do require special tooling to even produce basic components. You're probably right that we're more than 20 years from self-replication ability (again, not practicality), but I'd be surprised if it's more than 50 years out.
[1] https://arc.aiaa.org/doi/abs/10.2514/1.A33409 [2] https://3dprintingindustry.com/research/
