Calling fsyncLock multiple times without calling fsyncUnlock in between can cause a race condition and block a node until the process is restarted, see SERVER-6302.

Until the bug is fixed, users should be warned in the documentation for fsyncLock and fsyncUnlock that this can happen and advise against calling fsyncLock repeatedly. A work-around (inspecting if a node is locked before locking) is given on the server ticket.

A common use case is to run automated backup scripts repeatedly (e.g. cron jobs). If the backup doesn't finish before the next one is scheduled, these kinds of nested calls can happen.