A company is migrating its massive multiplayer game's infrastructure to AWS. The leaderboard needs microsecond read latencies and single-digit-millisecond write latencies, the dataset size is single-digit terabytes, and writes must become available within a minute after a primary node failure. Data must also be persisted for later analytics. Which option meets these requirements with the least operational overhead?

Amazon MemoryDB for Redis in Multi-AZ mode is the best choice because it natively provides microsecond read and single-digit-millisecond write latencies, crucial for a game leaderboard. Its Multi-AZ deployment ensures high availability and automatic failover within a minute, meeting the RTO requirement. MemoryDB is also durable, persisting data to Amazon S3 for analytics without additional configuration, and is a fully managed service, minimizing operational overhead. Amazon RDS is not suitable because it cannot consistently achieve microsecond read latencies. While it offers durability and managed services, its performance profile doesn't align with the strict latency requirements. Running Redis on Amazon EC2 instances would require significant operational overhead for setup, patching, backups, scaling, and high availability, which contradicts the "least operational overhead" requirement.