TST: Add regression test for MultiIndex merge after reset_index

pandas/tests/reshape/merge/test_merge.py

@@ -2936,6 +2936,44 @@ def test_merge_multiindex_single_level():
     tm.assert_frame_equal(result, expected)
 
 
+def test_merge_multiindex_reset_index_mixed():
+    # GH#62150 - merging reset MultiIndex when one df has single-level index
+    # Bug existed in 2.3.1, fixed in 3.0.0, this test prevents regression
+
+    # Create first dataframe with MultiIndex on both index and columns
+    df = DataFrame(data={("column_1", ""): [1, 1], ("column_2", ""): [2, 2]})
+    df.index = MultiIndex.from_arrays(
+        [[1, 1], ["metadata_1", "metadata_2"]], names=["index", "metadata"]
+    )
+
+    # Create second dataframe with single index and MultiIndex columns
+    df2 = DataFrame(data=[1, 1], index=[1, 1]).rename_axis("index", axis=0)
+    df2.columns = MultiIndex.from_product([["new_data"], [""]])
+
+    # Test merge with on='index' - should work
+    # Note: PerformanceWarning is expected due to non-lexsorted MultiIndex
+    with tm.assert_produces_warning(pd.errors.PerformanceWarning):
+        result = df.reset_index().merge(df2.reset_index(), on="index")
+
+    expected = DataFrame(
+        {
+            ("index", ""): [1, 1, 1, 1],
+            ("metadata", ""): ["metadata_1", "metadata_1", "metadata_2", "metadata_2"],
+            ("column_1", ""): [1, 1, 1, 1],
+            ("column_2", ""): [2, 2, 2, 2],
+            ("new_data", ""): [1, 1, 1, 1],
+        }
+    )
+    expected.columns = MultiIndex.from_tuples(expected.columns)
+
+    tm.assert_frame_equal(result, expected)
+
+    # Test merge with on=[('index', '')] - should also work
+    # Warning may or may not be raised depending on internal state
+    result2 = df.reset_index().merge(df2.reset_index(), on=[("index", "")])
+    tm.assert_frame_equal(result2, expected)
+
+
 @pytest.mark.parametrize("on_index", [True, False])
 @pytest.mark.parametrize("left_unique", [True, False])
 @pytest.mark.parametrize("left_monotonic", [True, False])