If during replication recovery, the time it takes to apply a batch of oplog entries is greater than the cursor timeout period, the system will crash. We should set the cursor (which is local) to kNoTimeout to avoid this.