[ryanmcdonough]

Now, I may well be naive - but isn't the point of these systems that you fail over gracefully to another data centre and no-one notices?

[codeulike]

I get the impression that this has been thought about to some extent, but its a constantly changing architecture with new layers and new ideas being added, so for every bit of progress there's the chance of new Single Points Of Failure being added. This time it seems to be a DNS problem with DynamoDB