Had a similar problem months ago with some server in the US. This sort of issue should be detected in minutes and users re-routed. Perhaps the amount of traffic is limiting this as an option.
I also noticed it started late yesterday to a number of US sites. I haven't had much issue with Canadian sites, but when I go to the US sites... I often have issues. Guess it depends on the routing (may not be US/Canada but that was a trend for me).
I haven't bothered running a traceroute, but that should track the issue.