The oplog_lag_secs metric is only set to a non-zero value after the first flush succeeds. We should have a way to detect when the initial flush is being held up.

Some ideas:

• Log time since last completed flush

We should also consider logging the persisted valid-as-of timestamp, with this value being unset until the first flush. This would also let us tell at a glance what the last persisted valid-as-of timestamp is set to, without needing to read from containers.