This issue is connected with https://sourceforge.net/tracker/?func=detail&aid=2871316&group_id=96251&atid=614171
Some pages on the wild web contains comment start inside an literal mode element without corresponding closing sequence.
Browsers (I've checked FireFox and Lynx) display the body of the page b.html (see example) correctly.
The tidy utility sees that page as a perfectly legal page.
// > tidy -e -utf8 b.html
// line 1 column 1 - Warning: missing <!DOCTYPE> declaration
// Info: Document content looks like HTML 4.01 Transitional
// Info: No system identifier in emitted doctype
// 1 warning, 0 errors were found!
Example with comment start in literal mode element.