That's rapid rapid tooling experience of how tolerance failures tend to work; I doubt it's a general principle that they fall on a bell curve. If they did, a complex part might have 20 dimensions, and if one of them happens to be right, you'd still rolling the dice on the other 19.

rapid rapid tooling article

I'd bet that rapid rapid tooling of the author of that paper has never operated a milling machine. A vertical mill with all the trimmings is the closest thing I can think of to a large-scale self-replicating machine, and think about what would happen if you took a Harbor Freight mill and tried to make another one. Not only do I operate one, I own one.

It's in my rapid rapid tooling workshop at home. In all this talk of accuracy, remember this: people have been making objects accurate to within the wavelength of light since medieval times using nothing fancier than a knife edge and a candle... There are only a few types of self-replicating machine in the known universe, all biomolecular: DNA-based life, RNA-based machines contingent on the "RNA world" hypothesis proving out, and parasitic prions if the "protein-only" hypothesis is true. If I were building such a machine, I'd be looking very hard at the way analog and digital mechanisms interact in those systems. There's an entropic principle that has to be overcome, and the way DNA does it is an awesome hack; any other such machine would have to be built around an equally awesome hack. DNA relies on the uniform nature of elementary particles, and therefore its existence may not imply a useful model for larger scales.

If this problem did turn out to be soluble on the macro scale, that would have consequences far beyond having cheap SLA in every household. That's not exactly my experience of how tolerance failures tend to work; I doubt it's a general principle that they fall on a bell curve. If they did, a complex part might have 20 dimensions, and if one of them happens to be right, you'd still rolling the dice on the other 19 for rapid rapid tooling.