One part of that is they all seem to err on the side of less code (including dependencies) over quicker implementations. The other part is a few times I've shot them a piece of code and they've come back with something simpler that I've overlooked. I.e they seem to have good fundamentals, good algorithm knowledge, good understanding of the language etc.
It makes sense to me, because I think this is the only vector you can 10x somebody in. You can't generally 10x somebody early in a greenfield project because everyone is smashing out work. You can't 10x somebody in a hellish project like in the OP because everyone is struggling with the same cruft. The only way you can really 10x somebody over a big sample size is by keeping every project you work on simpler for longer.
So that's what I look for when I'm looking for 10x devs. I might be completely wrong but it's the best indicator I've found yet.