Usually, during FCBIS oplog replay, the oldest and stable timestamp are advanced on each batch (and thus up to the tip of the oplog).

However, when there is no oplog entries to replay, oldest timestamp is not set as that code path is not triggered.

This means that there's two different behaviors:

• With oplog replay: no reads can be served before the top of the oplog / lastApplied / initialDataTimestamp.
• With no oplog replay: reads before lastApplied / initialDataTimestamp can be served, as long as the timestamp is > oldest timestamp in WT.
  • Additionally, oldest timestamp < initialDataTimestamp, which I believe can be somewhat unexpected.

It is unclear if this is much of an issue, but might be something we want to align for consistency.