What helped me a lot with pushing deeper down into the language innards is to have people to explain things to.

Last week, for example, one of our students asked what closures are.
Explaining that was no problem, I was also able to differentiate them from function pointers, but then she asked what in Rust the traits/interfaces Fn, FnMut and FnOnce did (which are implemented by different closures).

And yep, she struck right into a blank spot of my knowledge with that.
I have enough of an idea of them to just fill in something and let the compiler tell me off when I did it wrong.
Even when designing an API, I've worked out that you should start with an FnOnce and only progress to FnMut, then Fn and then a function pointer, as the compiler shouts at you (basically they're more specific and more restrictive for what the implementer of the closure is allowed to do).

But yeah, these rules of thumb just don't suffice for an actual explanation.
I couldn't tell you why these different traits are necessary or what the precise differences are.
So, we've been learning about them together and I have a much better understanding now.

Even in terms of closures in general (independent of the language), where I thought I had a pretty good idea, I had the epiphany that closures have two ways of providing parameters, one for the implementer (captured out of the context) and one for the caller (parameter list).
Obviously, I was aware of that on some level, as I had been using it plenty times, but I never had as clear of an idea of it before.