database replication lag. how do you detect and handle it? dont you hate when it happens? i know i do.

I've been pondering replication lag in distributed databases lately. We use replicas for high availability and read scaling, right? But then, what happens when they lag behind the main one? It's easy to brush it off until it actually causes problems, like people seeing old data or transactions suddenly failing. So, how are you all keeping an eye on replication lag *before* it becomes a headache? And what do you do to lessen the blow when it does happen? Are there particular tools or methods you trust to keep data consistent, even with lag?