Cache Coherency in Geographically Distributed Microservices: Active-Active vs. Read-Through?

Hey everyone, I'm trying to figure out a design for a new service. It needs really fast reads all over the world, even with data that changes a lot. We're thinking about putting it in different regions using microservices. The big issue is how to keep caches consistent without making writes or reads super slow. Has anyone actually set up active-active caching across regions? What kind of problems did you run into with conflicts or network delays? Or, on the flip side, has anyone managed to scale a read-through cache with strong eventual consistency that actually felt solid in a real product? What did you have to give up, and what did you learn about how to invalidate stuff in that kind of situation?