I also find that my "context windows" can only hold so much. I can't understand everything all at once. So if I do very deep dives, I tend to forget (temporarily) other parts of the problem/stack/task/whatever. The solution for me is to create as much as possible as self-contained testable units. Not always possible, but something to strive for.

And I find that this skill of organising is the limit for how large/complex systems I can build before they become unmaintable. This limit has increased over time, thankfully.