Reporting of commands which invoke further processing in currentOp() could be more intuitive, for example extra description could be added to sub-ops. To reproduce:

mongo <mongod>
> while( true ){ db.findAndModify({}); sleep(1) }

mongo <mongod>
> while( true )

{ printjson( db.currentOp() ); sleep(1) }

Sub-ops reported in various completion states, no way to know what the root cause was. Repeated GLE can also result in empty currentOp()s.