I don't agree either, I think it's _not a flaw_, not a virtue! Anyway, all I talk about is the _writing style_, not the content of a scientific paper. That's a different thing. But you can't really teach how to make exciting research contributions.
I agree also that repetition is not a flaw in and of itself. Minimalism is a thing, after all.
As to older papers, I kind of agree that they are often more pleasant to read, but the problem is that the more people work on a subject, the more new, subtle aspects of the subject become apparent that have to be made clear in writing.
Also, this is very unfortunate but terms can become overloaded with time as people use them in different ways. Sometimes I read older papers that seem to take some term for granted, but I find myself being utterly confused about what, exactly, they mean, among many different meanings of the term. For example, some of the reviewers I note above were frustrated with my work complained about my use of the word "predicate", which I was certain at the time can't be misunderstood... but is in fact used in different ways by different communities.
Thanks for the interest in my work. My second Machine Learning Journal paper is the much better written one, although it's rather a bit thick with formalisms and proofs etc. and won't be easy to read without a background in logic programming.
I've also run into the same terminological overloading problem. In CS it's especially bad; papers from the 01980s would be already difficult to read for someone who'd only read the recent literature. The worst is when the different meanings are almost, but not quite, the same. For example, I was reading one of Steinmetz's first papers from the 01890s, and he uses the term "inductance" where now we would use "inductive reactance"; at a given frequency these quantities are proportional but the second one varies by frequency.
My knowledge of logic programming is not literally zero in that I've written Prolog code a few times and ported μKanren to OCaml in order to understand it, but it's also far from a deep understanding. Still, I'm optimistic about my ability to understand things.
Explaining and clarifying are not in conflict with being interesting, except in a few marginal cases like this one. And repetition of terminology is not inherently boring—the guy who wrote "κομόωντας Ἀχαιοὺς" 26 times and "γλαυκῶπις Ἀθήνη" 32 times would certainly dissent from that, and the Tipitaka or any Beatles song is far more repetitive still—it's just that terminological variation is one possible way to add interest. A bad one, in this context.
When I read scientific papers from a hundred years ago, before the institution of peer review, I'm constantly struck by how much better written they are than current papers. They talk about things that surprised them and disappointments that happened during the experiment, and their choice of words is much more lyrical and even metaphorical than the current literature. Consequently they are much clearer and explain much better. The universal complaint about the modern literature is that nobody reads it, even when they cite it, and that is only a slight exaggeration; one reason for this is that reading it is so painful, so people devote less time to it than they would were it more pleasant.
So, I agree with you that current academic writing norms prohibit publishing well-written, interesting prose; where we differ is that I think that is bad.
For context, my experience publishing peer-reviewed research is not quite zero, but you have much more than I do. I haven't read your papers yet (though they look interesting), so I'm not attacking your papers as dull or badly written. It's possible I've read more research than you have just as a result of being much older.
On the other hand, it's also true that many of those old papers included fatal ambiguities which the modern process would have weeded out. Still, we can have good-quality, interesting writing without ambiguity.