Currently, if we fail to sample the oplog for some reason, we fallback to scanning the entire oplog. Now that this process has been moved to the asynchronous cap maintainer thread, ensure that it still works as expected.

Some things to consider:

• Previously a collection scan wouldn't encounter things like cache pressure, but as a workload can be running concurrently now and filling the cache, do we need to ensure that we handle things like cache pressure?
• Does the collection scan have to yield to be able to see more oplog data as the oplog is getting written to in parallel?

This ticket will aim to investigate the answers to these considerations and any others that come up in relation.