I generally share your criticism, but my overall experience has been positive. Dask indeed requires some fiddling, and it's API and documentation has some blind

I generally share your criticism, but my overall experience has been positive. Dask indeed requires some fiddling, and it's API and documentation has some blind spots, but I've successfully used it to scale up some Pandas code to run on dataframes measuring in the tens of GB. Being able to transfer my knowledge of Pandas and NumPy so readily makes Dask's speed bumps worth it.