This is great research. Thanks for sharing it.

Of note is our recent analysis of longdesc behavior – http://webaim.org/techniques/alttext/longdesctestcases I have not changed the language on the survey results page based on these finding. In all cases where longdesc is supported (which is rather rare to begin with), if a longdesc references an in-page anchor it treats it identical to an external URL which means that the target page is opened in new tab or window. With an in-page anchor, it will set focus to that anchor in the new tab/window. So it would be a stretch to suggest that the longdesc is “optional text, available on the same page” when it is neither optional (only presented on demand) or treated as being on the same page.

PS – The default text in your Comment text box is super tiny.