Your question is Distributed System High Availability. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How do you ensure high availability in a distributed system?
Answer this in the context of a production ML service. Explain how you would design for failures across service instances, regions, dependencies, data stores, feature stores, and model-serving infrastructure. Cover redundancy, health checks, failover, graceful degradation, recovery, observability, and validation of the availability design.