Well I am not talking about a 1M node outage. That you cannot fix with anything. I am talking about a maximum datacenter wide outage, that actually happens pretty often. Amazon has game days, Netflix has chaos monkeys for the same reason. Make sure that you can rebuild parts of your infra pretty quick.
As long as your VMs are prebaked and your cloud provider supports ASG-esq primatives (i.e. "I need X instances running at a time, instantiate with such and such meta data") anyone can rebuild their prod infra quickly. You don't need Docker or containers to do that.
Oh, of course. Our datacenters often go offline (both planned and unplanned), and we are always ready to handle that. We are pretty much constantly re-provisioning servers... with so many physical machines, hard drive and other hardware failures are a daily occurrence.
