Replication lag is missing in Fly Dashboard

[davissp14]

[guillaumervls]

@davissp14 to help us get a sense of priority on this. How is this metric used? Are there frequent cases where having this metric would have prevented users to hit walls?

[elliotdickison]

I'll +1 this with a use-case: we alarm on high replication lag to give us an early warning if a replica is falling behind. This metric is important because a laggy replica often appears totally fine on every other technical point (normal memory, cpu, etc.), but totally wrong to users (why won't my data save? why can't I access new values? etc.) A few months after this metric stopped working we had a replica fall way behind and cause mass user pain in a region. We didn't catch the issue until we noticed a massive customer contact spike (not an ideal approach to error detection).