TypeError: '(slice(None, None, None), 0)' is an invalid key

[scidam]

I would suggest to put some print functions inside generic_combined_scorer., e.g.

def generic_combined_scorer(x1, o1, ii_1, x2, o2, ii_2, y, h):
        print(o1.columns, o2.columns etc.)  # print all columns names of df's etc
        s1 = h(x1, y)
        s2 = h(x2, y)
        o1[ii_1] = s1
        o2[ii_2] = s2

Don't pass arguments with indicies, e.g. X_subset.iloc[:, ii] could be replaced with X_subset, ii
is already passed, so you can do X_subset.iloc[:, ii] inside generic_combined_scorer. This should simplify debugging.