pandas drop_duplicates: keep='first', and the Row It Chose For You

duplicated and drop_duplicates on the same eight rows: 7, 5 or 2 rows out depending on the arguments. Which row survives is decided by the sort, and pandas disagrees with Excel about case.