I think they mean “plain old semantic HTML” in which case it does pose an accessibility issue for some screen readers. It’s reasonable for a screen reader user to expect that when they hear “block quote [text] out of block quote” that [text] is a quote (be it another document, a book excerpt, something someone said), not something else. In the reverse, it would seem surprising that a block quote is quoting itself!
It does gracefully degrade for sighted users, but at the expense of the visually impaired, which I don’t think is an acceptable trade off. I’ve already encountered markdown-rendered documents in the wild that don’t properly transform “admonition block quotes” to regular divs.
The GitHub renderer doesn’t have this problem, but hiding admonitions away in block quotes and standardizing on it means that, for the next few years, a bunch of visually impaired users are going to be hearing confusing “block quote [text that isn't actually a quote] out of block quote” for things they were expecting to be block quotes.