In my undergrad we sent all our essays etc as LaTeX and it was honestly very usable. I assume this was somewhat dependent on having a wealth of enthusiast professors and postgrads to develop the templates though...?
Since then, I've used LaTeX in a "freeform" way and absolutely hated it, I will definitely be trying Typst next time I need more than Markdown/RST.
But I can imagine that if you are unlucky, working with the journal-provided templates could be WORSE - if the templates suck you are surely in a world of pain!
When you're on your own, you need to either steal someone else's and modify it, learn to build your own, or just use something like the Memoir package (which is the first but designed for it.)
Even if you don't use LaTeX at all, "A Few Notes on Book Design" is worth the read: https://ctan.org/pkg/memdesign
The process of writing in latex for journals (and textbooks -- I've written for both things) is really very easy. Publishers almost always provide a sample .tex file that has items you just fill in. For example, there will be something like `\authorNames{}` and you just put the names between the braces. The same goes for titles, equations, figures, etc. There will be sample paragraphs as well. And they will have examples of various citations styles, too.
Speaking of citations, latex has good support for citation databases. (Typst and markdown also provide this support.)
I'd say most people I know write their early drafts in latex. They have a target journal in mind from the start of the writing process, so they just grab the latest sample file and stylesheet(s) from the publisher's website and start entering text.
As you say, using latex for freeform documents is a different matter. That's where I've started to use typst. And I do recommend it for such things. You may find yourself wanting to make some typst templates for common tasks (meeting notes, position papers, etc.) but it's not terrible hard to make such templates. I've made a few, but often I don't bother -- I just put a few lines of customization at the start, if I want to alter how section headings look, or I want a different font from the default one.
Like other academics, I plan to stick with latex for journal articles and books, unless publishers provide support for typst.
Markdown still has a place for files on github, because that means that web browsers will display formatted material, not just the raw code. A similar thing applies to code documentation, with many systems (R, Julia, etc) supporting various flavours of markdown.
I recommend typst to students for small tasks like assignments. It offers more typesetting power than markdown, it's a great replacement for msword, and it's easier to learn than latex.
Long story short, typst fills a niche. But it's not the be-all and end-all, at least in my (natural science) field.