How would you design a multi-region, highly available architecture on AWS to survive a complete regional outage for Turing’s production workloads?
