Pandas: Sort innermost column group-wise based on other multilevel column excluding one row

Question

This is an extension to my previous question:

Consider below df:

In [68]: df = pd.DataFrame({\'A\': [\'a\'] * 11,
    ...: