Just realized that #HTML is extensible and #Markdown isn't.
HTML's #XML-like syntax allows for arbitrary non-standard tags that can be made sense of later, in the parser / scraper / renderer. Or animated with #JavaScript custom elements API. Note that I only mention XML syntax, not its semantics, with all the domains, schemes, namespaces, and imports of ontologies or whatever. I just talk random user-specific tags. That's more than enough for reliable ad-hoc extensibility if the syntax is uniform. And HTML syntax is uniform.
While Markdown syntax is not. It's arbitrary (and it doesn't consider the conventions that came before it: https://aartaka.me/email-formatting.html, but I digress) and thus extensions have to come up with random syntaxes for their custom entities. Which wont look good across Markdown renderers supporting the new syntax and the ones that don't. Especially given that new syntaxes (for whatever reason) always are quite ugly.
HTML will just treat new tags' contents as new blocks (legacy) or phrases (more modern), which is much more reliable than seeing a symbol soup in the rendered Markdown.