Your question is Consistency vs Availability in Distributed Systems. 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 balance consistency and availability in a distributed system?
Explain how you would make this tradeoff for an ML system with distributed feature storage, model serving, and asynchronous data pipelines. Discuss which operations require strong consistency, where eventual consistency is acceptable, and how the system should behave during partitions or stale-data conditions.