Based on several billion documents! Now that's what you call a good sample size.
Elliotte Harold counters the "considered harmful" party line:
This is false and a common myth perpetrated repeatedly by a few myopic programmers who can't see beyond the browser. I can and do do many different things with XHTML served as text/html that I cannot do with classic HTML. In particular I can use the full XML toolchain to process such documents.
Browser display is not the only purpose of HTML and XHTML documents on the web. This is especially true for WordPress, where plugins need to do all sorts of weird things with the pages, inserting, deleting, rearranging, and filtering content. This is far easier and more reliable if the pages are well-formed