You are really onto something here. I am in plain-old mechanical engineering. As a student, for the first years, I was simply intimidated. "These papers are so hard to read!". You have to read between the lines so much, fill in so many gaps, correct so many mental and grammatical knots.
10 years later and 2 years into a PhD, I think I am ready to admit that most papers are terrible. Most people are not native speakers. It makes a huge difference. Filler words, synonyms, long sentences, grammar errors... the list goes on. The only thing not on it are typos, because those are caught by software and are an easy fix. The rest takes genuine work and effort.
Papers are so hard to read, they can be headache-inducing and I am often completely exhausted from them.
And that's only touching on the stylistic part. Nowadays, everything is either all code or (strongly) relies on it (experimental work still needs collecting, evaluation, plotting, ...). But good luck finding actual, published source code (and if it's public, the sad-but-true adage of horrific academic coding takes effect). Methods are obscured and unclear. Sometimes, people can't even cite a single source properly almost verbatim -- fundamental reading comprehension seems to be lacking. Often, private data is used in private or public but unusable code, making papers entirely irreproducible. It's mental and frustrating. Coming across good papers in between the cruft really stands out, but happens at a low rate (10-20%?).
Most papers from before the 80s or so IME. After that, it seems like scientific papers became similar to patents: not written to elucidate, but to obscure and waste the time of readers. Readers are the competition in a big rat race, after all.
I find, in general, that philosophy tends to have the best overall quality of writing in academic writing. There are still examples of bad writing, but when it's good, the writing is crystal clear and a joy to read.
I found the 'Hallmarks of Cancer' papers by Weinberg and Vogelstein to be very well written. While those are reviews, their research articles are well written as well.
Given that attitude, your mileage may vary with these suggestions.
I utilize primary research all day every day for my job and my definition of good academic writing is clear, concise writing that enables me to answer the specific questions I am seeking to answers for quickly and in a manner that allows me to confidently cite the work (assuming the relevant information is present within the manuscript).
10 years later and 2 years into a PhD, I think I am ready to admit that most papers are terrible. Most people are not native speakers. It makes a huge difference. Filler words, synonyms, long sentences, grammar errors... the list goes on. The only thing not on it are typos, because those are caught by software and are an easy fix. The rest takes genuine work and effort.
Papers are so hard to read, they can be headache-inducing and I am often completely exhausted from them.
And that's only touching on the stylistic part. Nowadays, everything is either all code or (strongly) relies on it (experimental work still needs collecting, evaluation, plotting, ...). But good luck finding actual, published source code (and if it's public, the sad-but-true adage of horrific academic coding takes effect). Methods are obscured and unclear. Sometimes, people can't even cite a single source properly almost verbatim -- fundamental reading comprehension seems to be lacking. Often, private data is used in private or public but unusable code, making papers entirely irreproducible. It's mental and frustrating. Coming across good papers in between the cruft really stands out, but happens at a low rate (10-20%?).