Design a geo-distributed rate limiter

Design a rate limiting system that works across multiple geographic regions (e.g., us-east, eu-west, ap-south). Requirements: - Enforce per-API-key limits (e.g., 100 requests/sec) and possibly additional limits (per-IP, per-endpoint). - Requests can hit any region; the limit should be respected globally (not multiplied by number of regions). - Low latency in the request path; tolerate region failures. Follow-up: You have a worker system that must execute tasks under rate limiting, and you are considering two scheduling strategies: 1) Prioritize tasks with the largest cost/weight first. 2) Prioritize executing as many tasks as possible (maximize throughput / number of tasks). Explain pros/cons of each strategy and when you would choose which.