Is it a good idea to built fault tolerance over an Amazon SNS call. As Amazon services are built to be resilient should we add one more layer of fault tolerance or trust amazon to handle that part?
The fallacies are of distributed computing:
There are many network issues which can occur. Your application should also be resilient if network latency is high or even down on your side. It doesn't really matter which remote service you are calling.