I believe you do misunderstand. Availability Zones as you are referring to allow customers and AWS services to protect against EC2 and data center failures. They do not protect you against incidents within the software of AWS services. To take DynamoDB as an example, the outage in 2015 affected the entire region. It was not caused by a fault in EC2 or a data center. It was caused by a problem in the DynamoDB service itself. There is no protection you can currently give yourself against incidents like that, other than region-based failover. If there were multiple independent deployments of DynamoDB in a region, this would give AZ-like separation for the DynamoDB service, and allow customers to better protect against incidents that originate in DynamoDB.