The dataframe, df is static and cached. But running the below snippet gives different results each time.
to_keep = df.filter((df.rank == 1) & (df.new_date <