Looking at Lukas Biewald post in Dolores Labs startled me a bit.
Do all of those companies that use the S3 API to store data have no DR plan for any downtime due to S3 problems?
no application and hardware is immune to fail over.
What happens if their DB server is down will they stop servicing their clients due to that down time?
I think that looking at any infrastructure we need to assume there is a risk when using it and we need some sort of even temporary backup.
I am sure it will be hard to queue the thousands of pictures that SmugMug is handling in those 7 hrs of down time but a solution should be in place to enable that. especially when solution like ParkSpace allow you to work on S3 API with local resources.
All i can say about those applications fail time is Naive attitude to System Architecture that borders with neglegence. if you want to make a stable system you need a stable platform design and implementation.
I think that the main problem is again the hype of Elastic Cloud Computing that promises to relieve you from the need for DR as this is part of the system they developed. but due to its immaturity a company setting upon been in the early adopters group should consider the instability and eventual trouble it may encounter from this immature solution. this calls for a DR solution to these resources.
Now that we know that this is the case i expect more companies using the Elastic Cloud Computing to be more aware of the instability and treat the resources as another hardware resource.