When a user finds themselves in an out-of-disk-space situation, recovering can be tricky: data can only be removed after startup is complete, but without provisioning extra storage, startup will likely again fail due to disk full as a result of replaying the oplog for recovery purposes.

Investigate whether shrinking the oplog at startup, potentially in combination with restarting as standalone and running compact on the oplog, would actually help users and would be worthwhile in terms of development cost.