<a download> can sometimes just navigate #7718

Somewhat related to #5548, and will probably also be fixed as part of moving downloads into the navigation algorithm, but this is a separate issue:

If you put <a href="https://example.com/" download>Click me</a> on a page, this will not download example.com, at least in Chrome and Firefox: http://boom-bath.glitch.me/download-example.html

The current spec does not allow this. It allows bailing out and doing nothing, but the "download a hyperlink" algorithm never gets to the navigation algorithm.