Emacs fill-paragraph not breaking lines where expected

I have the following HTML code (a list item). The content isn't important--the problem is the end of line 2.

<li>Yes, you can learn how to play piano without becoming a
  great n开发者_运维百科otation reader,
  however, <strong class="warning">you <em class="emphatic">will</em>
  have to acquire a <em class="emphatic">very</em>basic amount
  of notation reading skill</strong>.  But the extremely
  difficult task of honing your note reading skills that
  classical students are required to endure for years and years
  is <em class="emphatic">totally non-existant</em>as a
  requirement for playing non-classical piano.</li>

The command fill-paragraph (M-q) has been applied. I can't for the life of me figure out why a line break is being placed on the second line after "reader," since there's more space available on that line to put "however,". Another weird thing I've noticed is that when I delete and then reapply the tab characters on lines 4 and 5 (starting with "have" and "of" respectively), two space characters are automatically inserted as well, like so:

<li>Yes, you can learn how to play piano without becoming a
  great notation reader,
  however, <strong class="warning">you <em class="emphatic">will</em>
    have to acquire a <em class="emphatic">very</em>basic amount
    of notation reading skill</strong>.  But the extremely
  difficult task of honing your note reading skills that
  classical students are required to endure for years and years
  is <em class="emphatic">totally non-existant</em>as a
  requirement for playing non-classical piano.</li>

I don't know if this is some kind of clue or not. This doesn't happen with any of the other lines.

Is this just a bug, or does any experienced Emacs person know what might be going on here?

Thank you

This is intentional. Lines that start with an XML or SGML tag are paragraph separator lines. If Emacs broke the paragraph in such a way that the tag ended up at the start of a line, subsequent applications of fill-paragraph would stop at that line. This is to ensure that, for instance,

<p>a paragraph</p>
<!-- no blank line -->
<p>another paragraph</p>

does not turn into

<p>a paragraph</p> <!-- no blank line --> <p>another paragraph</p>

For the same reason, Emacs will not break a line after a period unless there are two or more spaces after the period, because it uses a double space to distinguish between a period that ends a sentence and a period that ends an abbreviation, and breaking a line after the period that ends an abbreviation would create an ambiguous situation.

Looks like a bug to me.

I was able to trim down your example to something like this:

<li>blabla
  blabla <b>some_long_text_here</b> <b>more_long_text_here</b>

If I remove a single character of text from it, fill-paragraph works as expected. Or if I add a chacter between the two consequtive <b> elements.