"[conn557392] killcursors: found 0 of 1" in Mongodb replication primary log

I am running a production mongodb replication with version 2.6 now.

Today I found out that the primary mongod instance keeps writing log:

[conn557392] killcursors: found 0 of 1

And I checked the db.serverStatus().metrics.cursor, there are indeed a great amount of timedout cursor as mentioned in this discussion.

My questions are

1. Because I set all my read logic as Secondary preferred, the primary suppose to be write only. Why it needs to kill the cursor? The cursor suppose to be only for read?

2. Why all the application services are not influenced even though there are more than half million of timedOut cursor? Do I suppose to take this seriously?

Thanks in advance!