I experience the running out of RAM problem that you describe as well and take basically the same approach to you. That's much closer to how I read code.
My issue really is just a tendency to keep reading even if I didn't actually parse the sentence properly. It seems to be a complexity threshold rather than a volume one.
I think it's just what happens when I try to read a text at the top of, or beyond, my reading level. I imagine that anyone who struggled in school with reading comprehension may have a similar experience, or perhaps its like that feeling when you read a garden path sentence[1].
I struggled to even recognize that I was facing this issue because I found reading very easy during school. Facing a text that was too difficult to instantly parse and understand was rare enough that I never had to develop the mental tools to deal with it.
So I do what you say, except it's much more deliberate than just "slow down" and "re-read". I try to break the concepts down into chunks, and focus on being able to recall specific things about each chunk. First I convince myself that each chunk makes sense in isolation, and then what its "contract" is (how it interacts with related concepts and fits into the bigger picture). When I can easily recall both of those in a coherent way, I'm ready to move on. I use a variety of ways to practice recall, depending on whatever's most ergonomic (Anki, study sheet, deliberate practice, etc.). Sometimes I find that I lack enough understanding of other concepts to really understand a chunk, so I have to go recursively do the same thing for other concepts.
So I see it as more of a memory problem and not "understanding". I don't really know what "understanding" means, to me it's all just memory--memorizing why something is true/makes sense (by relating it to stuff I already know/memorized), and then memorizing how that thing fits into other contexts. If I can't load that stuff quickly into my working memory, then I don't "understand" it.