The journal flusher has to perform a write to update the oplogTruncateAfterPoint once per journal flush (here). A cache eviction stall in this storage engine write can therefore prevent MongoDB from flushing the journal, which can introduce worse stalls in the system.