Optimizing cache hit ratio in a geo-distributed service: trade-offs for consistency?

Hey everyone, I'm working on a service that serves content globally, and we're seeing some inconsistent cache hit ratios across different regions. Our current setup uses a CDN and regional Redis caches. We're primarily concerned with read performance, but strong consistency isn't strictly required for all data. We're considering implementing a more advanced caching strategy, perhaps with a global cache or more aggressive regional invalidation. What are some effective patterns for improving cache hit rates in geo-distributed systems, especially when dealing with varying levels of data staleness acceptance? Are there specific tools or architectural approaches you've found successful for balancing performance and eventual consistency in this context?