At the end of the day this seems to me to be still anecdotal and not very testable. Would it be possible, given the code history of n number of developers + quality of the output code, to create a predictive model of a software engineer's 'greatness'? Seems daunting and fraught with potential problems but of a lot more value than just the typical platitudes; 'honesty, creativity, passion' (take three virtues, earthporn background etc)