The v1/(page|revision)/{title}/html endpoints curently get all HTML from ParsoidOutputAccess. ParsoidOutputAccess will fail hard for content models not supported by Parsoid (currently, wikitext and JSON).
For example, the following should not trigger a 500: https://www.wikidata.org/w/rest.php/v1/page/Q3/html.
We should return a 4xx error for now. 415 perhaps.
Note that for now, we just need to avoid a hard crash. Eventually, the endpoints should handle requests for non-wikitext content (or multi-slot pages) gracefully, by using ParserOutputAccess instead of ParsoidOutputAccess if the content model isn't supported by parsoid. See T311648: Allow WikitextContentHandler to use Parsoid for rendering wikitext.