A lot of noise recently, triggered by the recent S3 Outage, on the fragility of Web Services…
And guess what? This is (and will continue to be) a recurring and expected theme on the highly interdependent world of Mashups and services on the web.
When having such dependencies, you need to take the time to understand the robustness of the services/platform that you are depending on, and make sure there is a good Service Level Agreement (SLA) in place. And then, wish for the best, because web services do and will fail on you. So, if you have critical dependencies, you must do infrastructure and capacity planning (see below), and contemplate going with a redundant fail safe approach, but that is going to cost you extra (as in time and money).
Note that Amazon S3 does have what seems a reasonable SLA.
Related to this see:
ceo