we used ninja as a parallel task runner in pytype - had our whole-project analyser generate a ninja file with a task graph, and then just evoke ninja to run it,

we used ninja as a parallel task runner in pytype - had our whole-project analyser generate a ninja file with a task graph, and then just evoke ninja to run it, taking care of dependencies and parallel execution. it worked very nicely indeed.