Writing style is subjective and I find Knuth's delightful: almost every paper of his that I've read has been fun. Even in his technical papers he writes at the level an undergraduate student with decent mathematical background (or a strong high-schooler) could mostly understand, which suits me just fine, especially compared to some other academics who write only for their peers / graduate-level and higher. And in any case, all the mathematical parts of TAOCP can be safely skipped if one is not interested.

But regarding your complaint about formatting, note that the book design of the TAOCP volumes is not Knuth's doing: someone at Addison-Wesley came up with them in the mid-1960s, and the first three volumes were published with Monotype (hot-metal) typesetting. When Knuth wrote TeX in the late 70s/early 80s, his goal was simply to retain the style of the books. (The layout was fine; the publishers were trying to move to phototypesetting and the fonts were poor https://tex.stackexchange.com/questions/367058, so he needed digital fonts and wrote METAFONT to create them, and he needed TeX just to be able to use those fonts.) Perhaps the book Concrete Mathematics will be a better example for you to critique, as Knuth had more of a hand in its typesetting (though there too a book designer was involved; the article Typesetting Concrete Mathematics has some details, reprinted in the Digital Typography volume and a poor scan of its earlier publication here: https://tug.org/tugboat/tb10-1/tb23knut.pdf).

About chapter length: the initial table of contents was one book of 12 chapters; Knuth wildly underestimated the printed length of what he had written (not many people know he actually wrote the whole book, all 12 chapters, in the 1960s, amounting to about 3500 pages — since then you can say he's been updating the chapters and publishing each section when he thinks it's ready), so the plan changed to seven volumes with the same chapters. And the chapter sections' lengths have themselves grown with the growth of the subject; as you can see, the current volume 4B is just a tiny fraction of the projected Chapter 7: covers 7.2.2.1 and 7.2.2.2.

Also, about the flowchart for how to read the books: it's partly a joke about flowcharts, but the suggested reading order is very much linear: https://i.stack.imgur.com/UCsOx.png — in words, it's just something like "read the chapters in order, skipping starred sections on the first reading. Also, feel free to skip boring sections, and skim math when it gets difficult" (he addresses the mathematics in the preface on p. viii).