`as_numpy` changes MultiIndex

What happened?

I have a DataArray with a MultiIndex. In some cases, I use dask, so I call .as_numpy() to compute the array and store it in memory. I would expect that this call does NOT change the MultiIndex.

This is the case for .persist(), however, it's not the case for .as_numpy().

In the following MWE the original coordinates are:

Coordinates:
  * z        (z) int64 0 1 2 3 4
  * r        (r) object MultiIndex
  * x        (r) int64 0 0 0 1 1 1
  * y        (r) int64 0 1 2 0 1 2

After .persist() we get the same result. After .as_numpy() we get

Coordinates:
  * z        (z) int64 0 1 2 3 4
  * r        (r) object (0, 0) (0, 1) (0, 2) (1, 0) (1, 1) (1, 2)
  * x        (r) int64 0 0 0 1 1 1
  * y        (r) int64 0 1 2 0 1 2

which is not the same and can lead to issues down the line.

What did you expect to happen?

No response

Minimal Complete Verifiable Example

import xarray as xr

da = xr.DataArray(1, dims=('x', 'y', 'z'), coords={'x': [0, 1], 'y': [0, 1, 2], 'z': [0, 1, 2, 3, 4]})
da = da.stack(r=('x', 'y'))


xr.testing.assert_equal(da['r'], da.persist()['r'])   # ok
xr.testing.assert_equal(da['r'], da.as_numpy()['r'])  # err

xr.testing.assert_equal(da, da.persist())   # ok
xr.testing.assert_equal(da, da.as_numpy())  # ok but should err

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.

Relevant log output

No response

Anything else we need to know?

No response

Environment

Reproduce the MultiIndex example using DataArray.stack(), then compare the coordinate results from .persist() and .as_numpy(). Trace the .as_numpy() entry point and add a regression test showing that converting the data to NumPy preserves the original MultiIndex and remains consistent with .persist().