May be technology can help a bit here? I think the issue in this case was more about the granularity of "takedown". If there was some kind of an API contract between the infrastructure provider and the application service provider - an API that lets you shutdown a single subdomain, an email service and so on, the situation would be more tolerable.
Or even something simpler - call the account holder's phone and play a simple recorded message that says "hey, its Rackspace, we need you to deal with an urgent issue, please check your email ASAP". Cheap for Rackspace, gives the customer fair warning, and if there's no response within some reasonable time frame then shutdown the machines in question.
Well I didn't agree or disagree with one hour, I said "some reasonable time frame" explicitly because the expected response time needs to consider what the real risks are.
I keep my phone permanently on vibrate (so I don't have to remember to silence it when in class, etc.), and besides, it'd take a hell of a lot more to wake me up than a ringing phone.
