Redis cluster vs Redis Sentinel: when do you need a cluster?

We're running Redis Sentinel today for high availability for our caching layer, handling about 50,000 operations per second. We're anticipating traffic to grow to around 200,000 operations per second in the next year. The question is, at what point do we need to migrate to Redis Cluster? Sentinel handles failover fine, but a single Redis instance has its limits on CPU and network I/O. Redis Cluster adds sharding and distributed data, which is great for scaling reads and writes beyond a single node. But it also introduces complexity: hash slots, key routing, and operational overhead. Is 200,000 ops/sec still achievable on a highly optimized single Redis instance (or a primary with multiple read replicas for reads), or is that definitely a tipping point for migrating to a full cluster setup? I'm trying to weigh the operational simplicity of Sentinel against the raw scalability of Cluster, and understand when the benefits of the latter truly outweigh the added complexity.