Currently if the read concern specifies one of these two options, we may do a no-op write if needed. If the read is on a secondary, the node waits for a short period of time, and then requests the primary to write a no-op if the lastApplied still cannot catch-up to the specified afterClusterTime / atClusterTime after the wait. We will change this to wait for the lastWritten opTime to catch-up instead of lastApplied because to decide whether a no-op write is needed, we just need to know that the specified opTime exists in the oplog, without having it to be applied.