Concatenate 3D array with 2D array

The 3D array has three dims: z, y and x.
The 2D array has two dims: y and x.
When I try to concatenate them by expanding the 2D array with z dim, there's something wrong in _dataset_concat

MCVE Code Sample

import xarray as xr
import numpy as np

x = 2
y = 4
z = 3
data = np.arange(x*y*z).reshape(z, x, y)

# 3d array with coords
a = xr.DataArray(data, dims=['z', 'y', 'x'], coords={'z': np.arange(z)})

# 2d array without coords
b = xr.DataArray(np.arange(x*y).reshape(x, y)*1.5, dims=['y', 'x'])

# expand 2d to 3d
b = b.expand_dims('z')

# concat
comb = xr.concat([a, b], dim='z')

Expected Output

Same as np.concatenate:

concat_array = np.concatenate((a, b))
comb = xr.DataArray(concat_array, dims={'z', 'y', 'x'})

<xarray.DataArray (z: 4, x: 2, y: 4)>
array([[[ 0. ,  1. ,  2. ,  3. ],
        [ 4. ,  5. ,  6. ,  7. ]],

       [[ 8. ,  9. , 10. , 11. ],
        [12. , 13. , 14. , 15. ]],

       [[16. , 17. , 18. , 19. ],
        [20. , 21. , 22. , 23. ]],

       [[ 0. ,  1.5,  3. ,  4.5],
        [ 6. ,  7.5,  9. , 10.5]]])
Dimensions without coordinates: z, x, y

Problem Description

    comb = xr.concat([a, b], dim='z')
  File "E:\miniconda3\envs\satpy\lib\site-packages\xarray\core\concat.py", line 135, in concat
    return f(objs, dim, data_vars, coords, compat, positions, fill_value, join)
  File "E:\miniconda3\envs\satpy\lib\site-packages\xarray\core\concat.py", line 455, in _dataarray_concat
    join=join,
  File "E:\miniconda3\envs\satpy\lib\site-packages\xarray\core\concat.py", line 395, in _dataset_concat
    raise ValueError("%r is not present in all datasets." % k)
ValueError: 'z' is not present in all datasets.

As suggested by @dcherian, assigning the coordinate label by changing b = b.expand_dims('z') to b = b.expand_dims(z=[3]) makes it work.

Versions

Start with the MCVE and inspect xarray/core/concat.py, especially _dataset_concat, where the reported ValueError is raised. Confirm the behavior with the 3D and expanded 2D arrays, then ensure concatenation succeeds and matches the expected four-slice result without requiring an explicit z coordinate.