Last week Amazon A3 was down for 4 hours and made a lot of webmasters unhappy. It further proves that it is quite risky to design your site solely relying on A3 to provide the essential functionalities, at least for now.
One way to reduce the risk is to have a copy of the files, for example images, saved in your server and design a flag in your code to pull the file from your own server if A3 downtime was detected. The flag can be controlled by some parameter in a configuration file so it can be easily switched.
One might argu this defeats the purpose of using AWS storage since the load copies take up the space. But I believe the storage cost will worth it in the event of the A3 hiccups. By using A3 service when it’s up and running, you will still save the bandwidth when serving those files, which is a lot more expensive than the storage cost.