Whenever you're ready, walk me through how you'd approach this.
Tell me how you ensure data consistency across distributed systems you have built or maintained.
Explain how you identify the source of truth, coordinate writes and reads, handle retries and partial failures, and detect or repair divergence. Discuss the tradeoffs between strong and eventual consistency, including any ML feature, training, or serving systems you would consider.