It's a topography change for load balancing most likely anti-average, as long as they are stored in the same data center and his backend connection methods use connection pooling, timeouts aren't too short, and the connection method is asynchronous, there is no risk of additional desync, there should actually be less as CPUs shouldn't fall short on connection threads as often on the front end. As long as the connections are in the same data center there is no increased risk of cheating.