Ask HN: AWS is telling us using Python for ETL is an antipattern, true?

Me and my team happily use Python + Dask for various things, one of them is ETL jobs. Most of what we do with Python is reading data via ODBC drivers and bulk uploading it into AWS S3 and Redshift. Now AWS is telling our IT that this is an antipattern. It seems Glue or Talend would be the best practice. I am confused as to how is having version controlled Python code to do ETL an antipattern? I can read and manage code much better than UI's. I always thought of tools like Talend as good to haves, but not as necessary if you have a solid team that can do Python. What is your experience?