An automation script fails to connect to an internal server exactly 1 out of 2 times it is executed. This behavior is seen from different clients. Which networking device must be at fault?

If the client laptop were at fault, the failure would be isolated to that specific machine and would not be reproducible from different clients.

A router failure causes consistent connectivity loss to a destination, not a predictable alternating 50% failure rate, because routing decisions affect all traffic to a destination uniformly.

A switch failure causes complete loss of connectivity for the affected segment, not a predictable alternating failure pattern observed from multiple different source clients.

A load balancer using round-robin distribution sends requests alternately to each backend server in its pool. If one of two backend servers is down or unreachable, every second request will fail regardless of which client initiates it, producing the exact observed pattern - 1 failure out of every 2 executions from any client. This consistent 50% failure rate seen across different clients points directly to the load balancer forwarding to an unhealthy backend.