Replication Lag in Distributed Databases: How Much Can You Handle?

hey everyone im wrestling with replication lag in a distributed sql setup. weve seeing occasional delays between writes on the primary and them showing up on secondaries. for most of our use cases, a few seconds of lag is fine, but for some critical reads, we really need strong consistency. whats your strategies for managing or tolerating replication lag. do you have specific thresholds you wont cross, or do you rely on read-your-writes consistency patterns. curious to hear your real-world experiences and approaches