Scaling read-heavy microservices: when to choose database replication over application-level caching?

Hey everyone, I'm working on a service that's becoming quite read-heavy, handling millions of requests daily for mostly static or slowly changing data. We currently use a shared database and a Redis cache at the application layer. The database is starting to show some strain, and I'm wondering if we're hitting the limits of our current caching strategy, or if it's time to seriously consider database replication (read replicas) as the primary scaling mechanism for reads. What are your thoughts on the breakpoints for each approach? Are there specific scenarios where one clearly outperforms the other in terms of cost, complexity, and performance?