The scenario described is typical of read-intensive workloads. A read-intensive workload will benefit from read replicas, which allow the workload to scale horizontally for read operations and handle a higher volume of concurrent read requests efficiently without impacting the performance of the primary database. Since most of the operations involve users reading content as opposed to writing, having multiple read replicas will ensure that the system can handle the high demand for read operations expected from millions of users. Multi-AZ deployment, while providing high availability, does not specifically address the read scalability issue. Sharding could help with write-intensive workloads by distributing data across multiple databases, but it doesn't address the read performance as effectively as read replicas. Write-through caching doesn't directly address the read-heavy nature of the workload, as it is typically used to ensure data consistency between a cache and a backend store, and is more related to write operations.