No, they don't. It depends on the MIME type. The spec says application/xhtml+xml pages MUST be parsed with draconian error handling, but pages with text/html need not. Note that text/html is deprecated for XHTML 1.0 but invalid for XHTML 1.1.
In other words, if we use XHTML 1.0 Transitional served with MIME type text/html we shouldn't run into problems.
Of course, we're already scrubbing users' HTML, so I don't see why it would be too difficult to clean it using f.ex the tagsoup algorithm.
Makeshifts last the longest.