Test run does exactly what you are looking at

Clicking Test run on a workflow whose steps existed only as an interview draft produced a success toast and then nothing. The button ran the published version, the sidebar showed the draft, and nothing said the two were different.

Test run now takes an explicit source, and whatever the sidebar shows is what fires. A draft fire pins an immutable snapshot, labelled v1.1, v1.2 and so on, so testing a draft can never publish it by accident. Testing the same draft twice reuses the snapshot rather than piling up history.

The check happens at the click. An empty set of instructions comes back as a readable refusal while you are still standing there, instead of a queue item that dies minutes later somewhere you cannot see.

A trigger that fails before a run exists keeps its reason and shows up in the queue as a failure you can open. Scheduled and webhook triggers that die that way also notify you, because nobody is watching a screen for those. A manual click already has you in front of it, so it does not.