Not Spaghetti

Seems to work OK.

Re: normalizing papers
That's not the way I'd put it.

1) I assume one has some audience in mind, meaning a distribution of knowledge.

2) A paper can either pick some range within that distribution, and it is always easiest to pick a narrow range and target that. Of course, if the target is "expert" means that many readers may not get beyond the title, as in the current issue of Science, in which I find:

The Hallucinogen N,N-Dimethyltryptamine (DMT) Is an Endogenous Sigma-1 Receptor Regulator

3) Or it may be that at least the abstract and intro cover a wider audience, and then it gets deep.

4) Finally, (and I think this is hard, and probably only appropriate for longer works), if one is writing for a broader audience range, the most disconcerting thing is to be reading along at a general level, and then suddenly, without warning dive to deeper level that seems hard to skip, as later (but still general sections) refer back to it

A good counter-example, where it's done well, is Hennessy & Patterson's "Computer Architecture - A quantitative approach". There one reads each chapter until the going gets hard, skips to the chapter summary, and then goes to the next chapter. Hence, a book used by serious computer architects can be recommended to some financial analysts.

5) I would think journals ought to be able to add value to this in terms of setting expectations for authors, and depending on the publication, helping the authors a little on this. Brian Kernighan & I once did a paper for IEEE Computer that we thought was already in pretty good shape, but one of their editors definitely helped improve it.