The parser is aware if HTMLTidy found errors or not. If HTMLTidy finds (and auto-corrects) errors, these pages should be auto-categorized. Presumably the HTML can be fixed by humans or bots or something.
I guess there could be gradations, depending on how specific HTMLTidy is. That is, Category:HTMLTidy found very serious errors; Category:HTMLTidy found mildly serious errors; etc.
Version: unspecified
Severity: enhancement
If bots can fix them, why aren't they also able to find them?
And if tidy auto-corrects them, why should humans bother "fixing" them?
(In reply to comment #1)
If bots can fix them, why aren't they also able to find them?
HTMLTidy is being run server-side. It's vastly more convenient for the server to auto-categorize the pages. Then a bot will be able to easily find them. :-)
And if tidy auto-corrects them, why should humans bother "fixing" them?
Graceful deprecation.
Some html errors are probably caused by MediaWiki itself, and not those pesky humans (I would guess. Esp. the table generation code. But I could be wrong)
There is an easy trick to test the output without HTML tidy applied: Use Special:ExpandTemplates. I did so with https://de.wikipedia.org/wiki/Spezial:Vorlagen_expandieren?wpInput={{Hilfe:Tabellen}} (en.wikipedia's help on tables is not transcludable), and validator.w3.org doesn't look like MediaWiki's table parsing generates broken HTML.