test: Fix test_failing_userdata_modules_exit_codes (#5279)

In f24ecef, we persist error state across mode invocations during
the same boot. Fix the integration testing accordingly.

tests/integration_tests/modules/test_cli.py

@@ -117,8 +117,18 @@ def test_invalid_userdata_schema(client: IntegrationInstance):
 
 @pytest.mark.user_data(FAILING_USER_DATA)
 def test_failing_userdata_modules_exit_codes(client: IntegrationInstance):
-    """Test failing in modules representd in exit status"""
+    """Test failing in modules representd in exit status.
+
+    To ensure we don't miss any errors or warnings if a service happens
+    to be restarted, any further module invocations will exit with error
+    on the same boot if a previous invocation exited with error.
+
+    In this test, both bootcmd and runcmd will exit with error the first time.
+    The second time, runcmd will run cleanly, but still exit with error.
+    Since bootcmd runs in init timeframe, and runcmd runs in final timeframe,
+    expect error from those two modes.
+    """
     for mode in ("init", "config", "final"):
         result = client.execute(f"cloud-init modules --mode {mode}")
-        assert result.failed if mode == "init" else result.ok
+        assert result.ok if mode == "config" else result.failed
         assert f"'modules:{mode}'" in result.stdout.strip()