[RECIPE FAILURE]: `recipe_autoassess_landsurface_soilmoisture.yml`: `AssertionError: ref and test must have same set of metrics`

[ehogan]

The RTW has identified that the recipe_autoassess_landsurface_soilmoisture.yml recipe fails to run:

Traceback (most recent call last):
  File "/opt/conda/envs/esmvaltool/lib/python3.13/site-packages/esmvaltool/diag_scripts/autoassess/plot_autoassess_metrics.py", line 104, in <module>
    main()
    ~~~~^^
  File "/opt/conda/envs/esmvaltool/lib/python3.13/site-packages/esmvaltool/diag_scripts/autoassess/plot_autoassess_metrics.py", line 83, in main
    plot_nac(
    ~~~~~~~~^
        control_model,
        ^^^^^^^^^^^^^^
    ...<10 lines>...
        config=cfg,
        ^^^^^^^^^^^
    )
    ^
  File "/opt/conda/envs/esmvaltool/lib/python3.13/site-packages/esmvaltool/diag_scripts/autoassess/_plot_mo_metrics.py", line 644, in plot_nac
    n_var = normalise(var, ref)
  File "/opt/conda/envs/esmvaltool/lib/python3.13/site-packages/esmvaltool/diag_scripts/autoassess/_plot_mo_metrics.py", line 425, in normalise
    assert sorted(test.keys()) == sorted(ref.keys()), (
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: ref and test must have same set of metrics

[alistairsellar]

I've found the cause of this error. #4085 enforced strict=True for all zip commands, and also for a local function normalise with its own strict kwarg that previously defaulted to False.

This zip is deliberately combining a list of test runs with a string of markers that can be longer than the number of tests

ESMValTool/esmvaltool/diag_scripts/autoassess/_plot_mo_metrics.py

Line 655 in 9262abb

for test, marker in zip(tests, MARKERS, strict=True):

So it should use a non-strict zip, but should have a comment explaining why, and also a check that the number of tests is not greater than the number of markers.

The local function normalise

ESMValTool/esmvaltool/diag_scripts/autoassess/_plot_mo_metrics.py

Line 412 in 9262abb

def normalise(test, ref, strict=True):

is intended to default to strict=False and I think was caught up in #4085 accidentally by automated reformatting.