This isn't anything new: Mozilla has
documented this "flaw" since 2000. The reason that this isn't that serious is that they can't retrieve an arbitrary history: just if you have visited a specific URL (not even a particular domain or site!) that it asks for. So unless the attacker explicitly asks if you've been to
http://example.org/47, it will never know that you've been to
http://example.org/47. It can't find that out by probing
http://example.org/,
http://example.org/47?x, or even
http://example.org/47/. It must probe
http://example.org/47, verbatim. Brute-forcing every possible URL? Yea, good luck with that. Brute-forcing every possible URL over a network connection? I think this just won the Most Impractical Exploit Award.
Practically, the extent of this flaw is to make for a nice sensationalist parlor trick for use by some grandstanding website that claims to have "discovered" a hole that people had known about for a decade, capable of determining if you've ever been to a common site, like google.com. Utterly useless for any sort of meaningful attack. Not to mention, it's slow and burdensome on the attacker's server.
As for disabling META refreshes, that's silly. It's not going to save you from this "exploit"--it's used by this particular
demo, and that's it. I can use this exploit to sniff your browser history regardless of whether you have disabled META refreshes. And disabling META refreshes in general is usually a pretty big hit on usability (for a selective block of META refreshes, you can use
NoRedirect, but with respect to this sniffing exploit, it's completely irrelevant).