Thanks for the post. Does your analysis remove forks before counting? If I recall correctly, Docker has tens of thousands of forks, enough to significantly skew any analysis that doesn’t compensate for it.