I think the picture you paint is too idealistic. Not many can read a scientific paper and understand exactly what it is trying to say, since we are trained to avoid adjectives, adverbs, and half-formed speculations when writing papers. What is omitted from a paper is often as illuminating as what is in the paper. To know what is not said, you must have an anticipation of what must have been said. Often, that is quite important in peer review.

Even when giving scientific talks, the advice I give my students and post-docs is to first have a clear idea of who the audience is: is it

a) experts only from your area? [e.g. conference]

b) people from the same department who may be generally aware of the area without knowing specifics? [e.g. job talk]

c) General public [e.g. showing lab to school kids. This is the level at which you talk to the public. soundbites - so it is invariably misrepresented by the media]

All three levels require different approaches for the same subject. (b) is especially tricky - you are talking to experts, but from a different field. So you have to simplify, while not being patronizing. As for (c), I think it is best to avoid it. Except in situations like the pandemic, it rarely is beneficial. Also, to do it right, it requires a lot of preparation.

That is fair, but I think the general skill of reading scientific papers is one that more people are capable of learning than are encouraged to learn. I certainly don't think anyone has the time or intellect to digest every paper that might be of interest to them. But a lot of people I've met seem to be interested in science beyond a 'headline' depth and simply don't think they have the capacity to read any papers at all. That seems to me like a shortfall that could be corrected with better public education and media behaviour.

I'd be curious if you disagree with any of that in theory or just have a better appreciation of the practical difficulties...

I agree with this. One more point is that journalistic summaries of scientific findings are flat out wrong. Journals like Quanta may represent a middle ground, which are readable yet sober.

I'm sure that many could read it, but reading it and understand are not the same thing. Understanding takes time, often a lot of time. And this is where the simplified bit comes back.

Discussions on social media rarely indicate much effort at nuance...

I'm not even confident about the "many could read it" part. Most papers are full of obscure jargon, and occasionally fields of study will just re-define words for their own purposes.