Design a globally scalable, low-latency distributed rate limiting system that can enforce fine-grained request quotas across millions of users and multiple microservices while ensuring high availability even during partial infrastructure failures.