Your question is Handle Service Failure 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 would you handle service failure in a distributed system?
Discuss the answer in the context of an ML-powered distributed service, covering dependency failures, degraded operation, recovery, and observability. Clarify assumptions before proposing an architecture, and explain how you would preserve correctness, availability, and safe model behavior during incidents.