nagios, possibly at noc02 (or both) should monitor mirrormanager metalinks and tell us if they start returning "Bad Request" instead of valid data.
We already had a check on /mirrorlist?repo=fedora-12&arch=i386&country=US but it did not trigger during the outage. My only assumption is that the f12 mirrorlist wasn't actually affected.
In puppet commit 0c419cf3 I have updated the check to look for rawhide instead of f12, changed arch to x86_64, and changed country to global. It now checks for the string 'development' as well, instead of 'releases' since rawhide's mirrors don't have the string 'releases'.