More client side retry built into protocols, and the ability to easily failover from the client-side. Protocols like DNS and SMTP are designed so that multiple

More client side retry built into protocols, and the ability to easily failover from the client-side. Protocols like DNS and SMTP are designed so that multiple servers that can handle traffic, and one going down isn't a big deal - the clients will just retry using and the entire system just keeps working. Compare to HTTP which don't have retry mechanisms which results in needing significant engineering to deal with single points of failure, fancy load balancers and other similar hard engineering challenges. Synchronization of state across HTTP requests would still be an issue, but that's already a problem in the load balancer case, pushed usually to a common backing database.