Design choice: Which rate-limiting algorithm would you choose (token bucket, leaky bucket, fixed window, sliding window, etc.) and why? Consistency vs availability: During a network partition, would you rather: Strictly enforce the limit (never exceed 100), or Allow slight overages to keep availability high? Explain your tradeoff using CAP theorem. Architecture: Sketch (in words) the system architecture: Where is the rate-limit state stored? How do regions coordinate (if at all)? What happens on each request? Edge case: How do you prevent a single user from bypassing limits by sending requests to multiple regions simultaneously? Failure handling: What happens if Redis (or your state store) goes down in one region?