Aurora Global Database has a single writer endpoint that always points to the primary region. If the primary region fails and the secondary is promoted, the writer endpoint updates to point to the new primary. However, during normal operations, writes to the global writer endpoint go to the primary region, which might be far away for users in us-west-2.
The latency for writes from us-west-2 to us-east-1 could be high. The design should consider using the local cluster for reads and the global writer for writes, but the write latency issue remains. The flaw is that writes from us-west-2 will always incur cross-region latency, which may be unacceptable.
A better design would be to use a multi-master or sharded database, but for Aurora Global Database, the only way to reduce write latency is to use the secondary cluster for writes only after failover. So the design flaw is that the write path is not optimized for the secondary region. However, among the options, the most significant flaw is likely that the application instances in us-west-2 are connecting to the global writer endpoint for writes, which routes to us-east-1, causing high latency.
But that is by design. The question asks for the most significant design flaw. Perhaps it's that there is no mechanism to redirect writes to the local cluster after failover without manual intervention? Actually, the global writer endpoint automatically updates, so that's fine.
The flaw might be that the application is using the global endpoint for reads? No, it uses local for reads. Maybe the flaw is that the health checks for ALBs are not configured to check the database health, so if the database fails, the ALB might still be healthy and traffic continues to a region with a failed database. That is a common design flaw: the application health check should include database connectivity.
So the most significant flaw is that the Route 53 health checks only check ALB health, not the overall application health including database.