Design a Concurrent Rate Limiter with Deferred Requests

Design a thread-safe rate limiter with the following requirements: Start with a single rate-limit rule, then extend the design to support multiple rules applied simultaneously. A rule can be represented as: at most limit requests may pass within a time window of length window. When a request is subject to multiple rules, it may execute only if every applicable rule allows it. A request handler may throw an exception. If handler execution fails, the request must not consume quota from any rule. Rate-limited requests must not be dropped. They should enter a waiting queue and be processed when capacity becomes available. Ensure correctness under concurrency: no rule may exceed its limit due to races, and no request may execute more than once. Discuss and address: Atomicity across checking and reserving quota for multiple rules; Quota rollback after handler failure; Queue fairness and head-of-line blocking; Out-of-order request completion; Window boundaries, monotonic clocks, and cleanup. Describe the core data structures, concurrency-control approach, request lifecycle, and time/space complexity.