I was in your position a year ago. I had used MicroPython extensively in projects and had embedded it in a larger project but had never really dived into the internals (i.e. what's an mp_obj_t, when to use MP_OBJ_FROM_PTR, etc).
One huge thing that helped was answering questions here (it's one of the main reasons I first started participating in the forum) -- many of the questions I answered required very specific and targeted spelunking that was highly educational. Often a lot of work to get a very simple answer. (That's not a dig at MicroPython, more that some of the questions involved some really subtle stuff).
I'd also echo Dave's suggestion about using GDB on the Unix port. I do that a lot...many times per week.
Another trick that's really useful is understanding the bytecode. Again using the Unix port, running it with "./micropython -v -v -v program.py" will dump the bytecode in readable form".
Some other things... there are a lot of clever and subtle tricks (often for code size) but for the most part MicroPython works the way you'd expect. The GC is very simple (and gc.c is entirely self-contained), it's a good starting point. (Here's a not very good explanation I gave a few months ago --
https://www.youtube.com/watch?v=H_xq8IYjh2w )
If you treat the NLR stuff like a black box, the VM dispatch is pretty neat too.
I think that as a C++ developer too, I took a while to get used to the macros everywhere, but I understand of course why they're necessary.
Where I get lost easily still though is in anything to do with the native emitters and the glue between them...
