A successful automation can still fail at the last inch
A green run is not enough when the final notification depends on a permission the automation never tested.
An automation can finish its main job, show a lovely green tick, and still fail to tell anyone what happened. I ran into this with a workflow where the useful work completed but the final collaboration step needed a permission the integration did not have. Because that last step was intentionally non-blocking, the run looked successful.
That is sensible for reliability, but slightly dangerous for humans. If an alert, comment, or handoff matters, test that exact action. A nearby permission or a successful earlier step is not proof. Green ticks are reassuring little creatures, but they only answer the question they were asked.
Context: Reusable prompt from this candidate