A big problem is prior knowledge. Most papers incrementally increase knowledge so by necessity they have to assume the reader knows it, lest they find themselves repeating it alot. How far back are you expected to go to help your audience understand?

The answer in most of academia is: not at all. You're expected to have learnt everything in the field up to that point. Academia is for academics and generally doesn't care about impact outside of academia, who seldom understand it anyway (because of its academic nature).

Every field has jargon and assumed knowledge. If you're writing a computer science paper, you're not going to start by teaching the reader grade school math.

Not necessarily. I agree, being a smart scientist doesn’t make one a good writer , however, most of the times the inaccessibility of a text comes from a high presumed knowledge of the reader. But if you dial the presumption down, then the readers versed in the field would have to drag through the text full of explanations they already know. And, excuse my assumption, most relevant citations come from the people in the field.