Thanks for sharing this solution!
Do you think it’s worth referencing this problem in the DataFrames Development Survey?