Yep, you're right that a single GET request has some problems. We don't right now (because we wanted to get feedback ASAP), but we will definitely be checking from multiple nodes (running on multiple different providers) by the time we're charging people money so that we're as sure as we can be that the problem isn't us.

Getting an API together (to enable a server daemon) is also very high on our priority list. (Although the disadvantage with that is that then you're not actually checking to see if the site's up: you're checking whether the server's alive, and that doesn't necessarily mean the site is up.) We do recognize we're dealing with something tricky and we want to get it right. Thanks for the feedback.