Last page of CHM incompletly rendered or missing

I have some trouble with the last pages of my documents.

In one case, i get some ERR011: Unable to parse HTML element on line 49! from htmldoc on the last pages.
The strange thing is, that if I re-run the very same htmldoc command dipalyed with the --verose --verbosity high level (obviosly not deleting the temporary files), the pdf will be complete and no error ERR011 is rised.
It's like the last file written by the conversion before invoking HTMLDOC is still open or not completely written!

In an other case, the last page is simply completly missing, without any error message.

But the code look just fine to me:

            pf=open(filename,'w')
            pf.write(page)
            pf.close

What could be wrong??

Writing a dummy file afterwards before invoking HTMLDOC resolves the problem, but this seems quite a ugly hack to me (I am not a programmer).

Then I found the command
           pf.flush()

This also solves the problem. But why is this necessary??