This is still the same behavior with #5692.

We would need to handle pd.RangeIndex (and perhaps range?) label indexers similarly to slice label indexers, i.e., use pd.Index.slice_indexer internally to return integer indexers as slices (*).

```python b = my_da.sel(time=slice(0, 2)) b.get_index('time')

RangeIndex(start=0, stop=3, step=1, name='time')

```

Otherwise, label indexers get internally converted to arrays. Note that the conversion to an Int64Index is done in pandas (nothing specific is done on the Xarray side), so I expect that this will be eventually addressed in pandas. This conversion may not be too problematic if we consider this as an implementation detail (although I might be missing some important aspect).

```python idx = pd.RangeIndex(0, 100)

idx[slice(0, 3)]

RangeIndex(start=0, stop=3, step=1)

idx[[0, 1, 2]]

Int64Index([0, 1, 2], dtype='int64')

```

(*) One major difference is that in Xarray slice label indexers are upper-bound inclusive, while pd.RangeIndex and range are not!