Why sharding cluster did not permit db.fsyncUnlock() command

I did not catch root of problem. I avoid this operation and use network restriction to manage income session to prevent write operation at maintenance event time. Problem is reproducible at my environment.