Are these not the 'friction can be ignored' assumptions of economics? They are, of course, blatantly false. But that doesn't stop such models from effectively modelling real-world behavior.
Granted, I know a slight bit about general equilibrium theory, but nothing about DSGE.
I'm always surprised when people suggest using a different language if you want typing in Python. Python's (second?) largest appeal is probably its extensive ecosystem. Whenever people suggest just changing languages, I wonder if they work in isolation, without the need for certain packages or co-worker proficiency in that language.
I understand your interpretation, and I agree with the first part of it. (Don't change an existing system without first understanding the implications of that change.') I think that's the core of the metaphore, as taken by most people.
I don't think the point is that you should never make changes to complex systems at all, though. I don't think its means that more primitive, or unaltered, states of a system are necessarily prefential to more altered states, which I infer from your comment.
If unalterated states were better, we would have to tear Chesterton's fence down — right? Fences don't occur naturally.