[daniel-cussen]

Not unlike the way floating point rounding errors accumulate, in fact. The only difference is you can mitigate those to a degree by increasing the digits of precision, whereas that can't really be done on a mechanical computer.

[asmithmd1]

You do that by increasing the physical size of the components

[jacquesm]

Continent size analog computer. Would be interesting to see if it would outperform IEEE doubles.