For a service deployed in two regions, how do you decide between a managed load balancer with cross-region routing and DNS-based failover? I’m comparing options like AWS Global Accelerator or Route 53, Azure Front Door, and Google Cloud Load Balancing.

DNS failover seems simpler, but I’m unsure how much client-side DNS caching can delay a switch after a region goes unhealthy. A load balancer may react faster, though it adds configuration and cost. How much do managed load balancer health checks help in practice, and what should be tested beyond whether the web server responds?

I’m especially interested in real-world tradeoffs when the app depends on a database or other regional services. For those setups, has DNS-based failover been reliable enough, or did you need a global routing layer?