AWS could setup multiple internal DNS paths to DynamoDB and change their DDB client libraries to fail over between them. That would have mitigated this problem, and that pattern could be applied to other critical services that underpin AWS functionality.