Multi-region database replication: our journey from single to global

we're planning to expand our platform to europe and asia, and the current single-region database architecture in the us is already showing latency issues for users in singapore (250ms+). we need a robust multi-region database strategy. we've considered traditional read replicas, but that doesn't solve the write latency problem. options like cockroachdb or yugabytedb offer geo-distributed capabilities, but they're a significant operational shift. another thought is using edge functions with regional data caches and a single primary, but that adds complexity to cache consistency. data residency requirements are also a factor. for those who've successfully moved from a single-region to a globally distributed database, what was your journey like, and what architectural patterns proved most effective?