AFAIK, Oplog is first stored in the secondary oplog and then applied by 16 replication threads.
The defaults are mostly good for almost all workloads so unless there is a support indication to tune this I would stick to defaults.
What is highly important to decide is the read and write concerns on the driver side as well as readPreference those will have a main impact on replication behaviour.
In the other DB (ex: Redis, etc), there was another space that went through before storing it in a log like oplog, so I was asking how MongoDB was doing.