Journal file preallocation and recycling (see preallocj in the code) is enabled based on the outcome of a short performance test run at process start. When mongod is deployed at scale this is likely to result in some servers using preallocj=true and others using preallocj=false. That non-determinism is not good for a DBA.
I get 10% better throughput from insert benchmark workloads using fast (PCIe flash) storage after forcing preallocj to be true. See http://smalldatum.blogspot.com/2014/03/redo-logs-in-mongodb-and-innodb.html