Did your benchmark for using pyarrow include transferring it to pandas?

Also you wrote at least 3 articles on avoiding CSV with pandas in September 2021 alone. This article is again a similar variation of the rest. With October do you intend to go through the remaining formats that pandas supports?