It's not the amount of "code".

They can program pretty competently and are learning miles faster than I thought they would - in fact all the software devs who join the company, who are themselves excellent programmers are consistently impressed at the code produced by the mechanical engineers.

It's the way that git is designed. The mental model needed to actually understand what is going on is a bit tricky, but git makes it massively harder than is needed by operating at multiple abstraction levels, all of which leak, and then lying to the users in error messages etc.