I was reading the Fast Data Blog and have a question about the implementation of loading the PyArrow tables, why is it that the implementation leverages a ThreadPool to fetch the files? Doesn’t

S3().get_many()

run in parallel as well?