You can see what's probably the same issue in Firefox/Chromium as well: You can't scroll e.g. https://tract-docs.dev/docs/getting-started/ with the keyboard without first clicking into the page.
Doesn’t seem fixed to me. Also, it’s impossible to fix it in the current structure without JavaScript, because you need to either move the caret into the desired scrolling element (`getSelection().collapse(document.querySelector('.main'))`), or focus it (`document.querySelector('.main').focus()`, also requiring tabindex=-1 or similar on the element). From the presence of tabindex=1 on .main, I’m guessing you tried that one.
For a documentation thing like this, I’d generally recommend avoiding a separate scrolling pane for the primary content, but rather arranging so that it can use the document’s scrolling element. This gets the scrollbar in the place people generally expect it in web things, and makes it work regardless of whether the JS loads.
I don’t know of any good articles about the trade-offs of the various approaches, but I’ve started writing one myself. (I’m drafting a series on the intricacies of various UI patterns and their implementations, and this was actually already on my list.) Email me if you’d like me to drop you a line when I publish it (being realistic, expect to wait weeks for it).
