Few will dispute that HTML5 will be adopted very differently than HTML. HTML was adopted by everyone at the same time because the W3 was the next big thing, browsers were crude and at first they all introduced their own "extensions". People (by which I mean me) authored HTML without even knowing what a doctype declaration was, and books that TAUGHT "web design" ignroed closing tags and semantics. That led people to author invalid documents in ignorance.

When W3C recomends their next version of (x)HTML, regardless whether well-formed, valid XML is required, the early adopters will be the highly technical people who author compliant XML by habit. The plebs will probably never learn the intricacies of HTML5. For them, HTML4 is good enough, and besides that the trend has been for sites to provide their own bracket-based markup, along the lines of wiki markup or bbcode.

So I wouldn't worry about messy HTML5 polluting the web... even if it is technically allowed I don't think many will use it.