I believe it's not. As a PaaS I would think that they would run a high availability cluster on at least 2 multiple regions so that they would have a mechanism

I believe it's not. As a PaaS I would think that they would run a high availability cluster on at least 2 multiple regions so that they would have a mechanism in place for events like these. I know it's expensive, but if you charge 250 for 2.5GB of RAM I believe you would have enough money to cover it. I also think as you hinted that they should separate services across different regions..