Markdown parser improperly handles certain HTML payloads

Using a naïve Markdown to HTML converter:

QTextDocument d;

d.setMarkdown(input); // Assume GFM with HTML enabled

std::cout << d.toHtml().toStdString() << std::endl;

The following HTML payload (which is also valid GFM):

<b>x&lt;</b>

gets converted to

&lt;<span style=" font-weight:600;">x</span>

(that's right, the escaped < ends up coming before the bold span). Meanwhile, the following payload:

<b>&amp;</b>

literally strips <b> (or whatever other element it would have been put to), becoming

&amp;

(boilerplate HTML before and after the key portion omitted).