https://developers.google.com/search/docs/guides/mark-up-lis...
The search results you see just put in the effort to provide structured data to google. You can do this easily and without any AMP integrations by simply embedding structured data (such as JSON-LD) on the page.
So there's no "effortless" way for knowledge of HTML-embedded microformats to filter down to most HTML document authors or web-app theme creators. They have to actually go and seek such knowledge out. And often they don't know what they don't know†, so they have no reason to.
Essentially anything on the web that requires or relies on an HTML-embedded microformat to work, is instantly relegated to being an "SEO black art" that just gets passed around in blog posts read by bigcorp web-app devs and no one else.
---
† This is one big part of what the people who dislike minification, WASM, etc. are complaining about re: the "readable" web. It used to be that you'd discover techniques like this by just looking at a bigcorp's production web pages and seeing how they work. That's far harder today than it used to be, and it's only going to get even harder in the future.
WebAssembly expressly has a textual representation so that it can be inspected, it's part of the spec. And for everything else that's compiled, there are always decompilers to give you the source, often times at great quality.
HTTP/2 is binary (and by default encrypted) but nobody notices because browsers and tools transparently convert it to text for you, even though there were plenty of complaints that it would be hard to inspect.
Can we be clear in this discussion that the oft-cited carousel is only relevant for trending/news type searches? It has no bearing on the overwhelming majority of searches.
Sites do it because they are forced to for that placement, not because they want to. They would much rather have a single consistent website that benefits everyone on the web.