Here are two ways someone commenting on a blog can subvert Firefox's feed sniffing into making the page unreadable:

Note that in both cases, the page remains valid HTML, and is served using the MIME type text/html, yet Firefox sniffs it as a feed. It's hard to argue that the blog's restrictions on HTML in comments have failed: in the first case the commenter has only included a simple command, and in the second case the commenter has only included text! The only thing that prevents this attack from working on most real-world blogs is that Firefox only sniffs the first 512 bytes.