Design a distributed system to enforce API rate limits across a global fleet of microservices. The system must support multi-tenant configurations, handle hundreds of thousands of requests per second with sub-millisecond overhead, and ensure high availability even during partial infrastructure failures.