Are numpy logical element wise operations broken for pandas 2.0? (np.logical_or)

I have a code that was working till I update to pandas 2.0, I checked in the changelog and I see that they changed the behavior of logical_or, but it doesn't seem exactly the same thing: https://github.com/pandas-dev/pandas/pull/37374 so it a is very unexpected error for me.

a = pd.DataFrame({"or":[False,True], "a":[True,True], "b":[True, False]})
np.logical_or(a[["a","b"]], a[2*["or"]])

Before 2.0 it use to do a element wise "or", now it directly fails 0.o with:

ValueError: cannot reindex on an axis with duplicate labels

Or if provided with two different labels in the second then is even worst as it doesn't fail but concatenates both in a extremely erratic way.

Is this a known bug on pandas / numpy? or is it intended? Is there any known efficient alternative?

The mentioned error is marked with # GH#42568 (https://github.com/pandas-dev/pandas/pull/42568/files).
Actually, the code they've changed is now checking/ensuring for unique labels in axes["columns"]. But, even if we change the labels on one of the input dataframes like

np.logical_or(a[["a","b"]], a[2*["or"]].set_axis(["d","c"], axis=1))

we'd get:

      a     b    c    d
0  True  True  NaN  NaN
1  True   NaN  NaN  NaN

as it seems like 2 array-like structures are expected to have the same column labels (if any) when compared:

In [642]: np.logical_or(a[["a","b"]], a[2*["or"]].set_axis(["a","b"], axis=1))
Out[642]: 
      a     b
0  True  True
1  True  True

A simple workaround is bypass labels check and pass only arrays of values:

In [644]: np.logical_or(a[["a","b"]].values, a[2*["or"]].values)
Out[644]: 
array([[ True,  True],
       [ True,  True]])