You are a Systems Engineer for a financial trading platform that requires extremely fast and reliable transaction ordering across multiple data centers worldwide. The platform must handle 100,000 transactions per second with sub-millisecond latency and can tolerate up to f simultaneous Byzantine failures out of 3f+1 replicas. Design, implement, and benchmark a BFT (Byzantine Fault Tolerant) consensus protocol optimized for this high-frequency trading scenario. Your solution should address the following: (1) Minimize communication complexity, (2) Reduce consensus latency in wide-area networks, (3) Optimize for throughput under normal conditions, (4) Provide fast recovery when replicas fail. Compare your approach against existing protocols like PBFT, HotStuff, and SBFT. Implement your solution in Go or Rust and provide performance benchmarks in a simulated environment that models network latency between different geographic regions. Analyze the trade-offs between consistency, availability, and partition tolerance under various failure scenarios. Include a discussion on how your protocol could be extended to handle dynamic membership and cross-shard transactions.