This entire thread is an epiphany for me and I can vouch your bit as something I share as well.
I used to get into loops with that - I thought I understood something, then I'd try to implement it and hit a corner case, or even a caveat would come to my mind, and suddenly my understanding is crashing. I developed anxiety around this, because thinking and rethinking would mean I could spend days without anything to show for it (other than a thousand lines of semi-coherent ramblings in an Org Mode file). I started to lean towards taking things simple and direct, solving them more by feel than explicit analysis, in order to create a bedrock of something I could then iteratively improve or rewrite, without having to keep the entire problem in my head at all times.
Yes.
I would have Monte Carlo simulation dreams that seem to prime my brain around the subject. But completely stupid and never based on the truth.
Not good sleep but I would seem to be more ‘tuned’ in on the subject.
"Understand it in my head" is often a good stopping point too. Because if i actually go and fix it, i may find out i was wrong and spend even more time re-analyzing the issue to reach a new understanding.