Neither the new HTML parser nor the old HTML parser has any explicit support for <xml>...</xml>. It's just invalid and unknown markup.
It would be more appropriate for developer.mozilla.org to present how to put XML content into an HTML5 "data block" (i.e. <script type="application/xml">...</script>) and parse it using DOMParser. This approach would work properly in Firefox, WebKit, Opera and IE9.
Do we have some kind of standard operating procedure for dealing with misleading docs? I'm a bit shy to just delete or overwrite the old doc on my own initiative.