PyTest random order for Fabric tests (#19040)

.github/workflows/ci-tests-fabric.yml

@@ -165,7 +165,7 @@ jobs:
         # NOTE: do not include coverage report here, see: https://github.com/nedbat/coveragepy/issues/1003
         run: |
           python -m coverage run --source ${{ env.COVERAGE_SCOPE }} \
-            -m pytest -v --timeout=30 --durations=50
+            -m pytest -v --timeout=30 --durations=50 --random-order-seed=$GITHUB_RUN_ID
 
       - name: Statistics
         if: success()

tests/tests_fabric/utilities/test_rank_zero.py

@@ -20,12 +20,12 @@ from lightning.fabric.utilities.rank_zero import _get_rank
         ({"JSM_NAMESPACE_RANK": "4"}, None),
     ],
 )
-def test_rank_zero_known_environment_variables(env_vars, expected):
+def test_rank_zero_known_environment_variables(env_vars, expected, monkeypatch):
     """Test that rank environment variables are properly checked for rank_zero_only."""
     with mock.patch.dict(os.environ, env_vars):
         # force module reload to re-trigger the rank_zero_only.rank global computation
-        sys.modules.pop("lightning_utilities.core.rank_zero", None)
-        sys.modules.pop("lightning.fabric.utilities.rank_zero", None)
+        monkeypatch.delitem(sys.modules, "lightning_utilities.core.rank_zero", raising=False)
+        monkeypatch.delitem(sys.modules, "lightning.fabric.utilities.rank_zero", raising=False)
         from lightning.fabric.utilities.rank_zero import rank_zero_only
 
         @rank_zero_only