BUG: Fix bug in DataFrame binary op not respecting fill_value in case…

pandas/core/frame.py

@@ -8029,10 +8029,15 @@ def _should_reindex_frame_op(self, right, op, axis: int, fill_value, level) -> b
             return False
 
         if (
-            isinstance(self.columns, MultiIndex)
-            or isinstance(right.columns, MultiIndex)
-        ) and not self.columns.equals(right.columns):
+            (
+                isinstance(self.columns, MultiIndex)
+                or isinstance(right.columns, MultiIndex)
+            )
+            and not self.columns.equals(right.columns)
+            and fill_value is None
+        ):
             # GH#60498 Reindex if MultiIndexe columns are not matching
+            # GH#60903 Don't reindex if fill_value is provided
             return True
 
         if fill_value is None and level is None and axis == 1:

pandas/tests/frame/test_arithmetic.py

@@ -2058,6 +2058,26 @@ def test_arithmetic_multiindex_column_align():
     tm.assert_frame_equal(result, expected)
 
 
+def test_arithmetic_multiindex_column_align_with_fillvalue():
+    # GH#60903
+    df1 = DataFrame(
+        data=[[1.0, 2.0]],
+        columns=MultiIndex.from_tuples([("A", "one"), ("A", "two")]),
+    )
+    df2 = DataFrame(
+        data=[[3.0, 4.0]],
+        columns=MultiIndex.from_tuples([("B", "one"), ("B", "two")]),
+    )
+    expected = DataFrame(
+        data=[[1.0, 2.0, 3.0, 4.0]],
+        columns=MultiIndex.from_tuples(
+            [("A", "one"), ("A", "two"), ("B", "one"), ("B", "two")]
+        ),
+    )
+    result = df1.add(df2, fill_value=0)
+    tm.assert_frame_equal(result, expected)
+
+
 def test_bool_frame_mult_float():
     # GH 18549
     df = DataFrame(True, list("ab"), list("cd"))