xr.concat error behaviour differs based on input DataArray ordering

What happened?

When using xr.concat without specifying the coords argument and with input DataArrays of differing dimensions, where one is missing a dimension that is present in the others, the behaviour of the function will depend on the order and number of the arrays in the input list. The expected behaviour is that a ValueError prompting the use of coords='minimal' will be raised, however this error is not raised when either:

• There are only two DataArrays in the provided list; or
• The DataArray with the missing dimension is first in the list.

In these instances, the xr.concat function acts as if coords='minimal' has been specified and broadcasts the smaller DataArray without erroring.

What did you expect to happen?

The raising of errors when using xr.concat should be consistent and not depend on the ordering or number of inputs.

Minimal Complete Verifiable Example

a = xr.DataArray([[1,2,3],[4,5,6]], coords=[('first', [10, 20]), ('second', ['a', 'b', 'c'])])
b = xr.DataArray([[10,20,30],[40,50,60]], coords=[('first', [30, 40]), ('second', ['a', 'b', 'c'])])
c = xr.DataArray([100, 200], coords=[('first', [10, 20])])

xr.concat([a, b, c], dim='first')  # Errors as expected
xr.concat([b, a, c], dim='first')  # Errors as expected
xr.concat([c, a, b], dim='first')  # Does not error
xr.concat([a, c], dim='first')     # Does not error
xr.concat([c, a], dim='first')     # Does not error

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

Start at the xr.concat entry point and reproduce the provided minimal example with each input ordering and list length. Trace how differing dimensions are handled when coords is omitted, then add regression coverage showing that all listed cases consistently raise the expected ValueError prompting coords='minimal'.