Is a workflow code or data?
The deepest question in the module, and it has an answer.
For most of its life one major engine treated a workflow as mutable data: you edited a row, and the active version was whatever the row said. That model cannot answer "what was running at 3am." It has now moved to a draft-and-published split, with production executions using the published version rather than your latest edits, and a version ID pinned on the execution.
A workflow is data at edit time and code at run time. The editor needs a mutable, autosaving document; the runtime needs an immutable, addressable artifact. Publishing is the transition, and the execution must pin the version it ran.
Get it wrong in either direction and you get one of two failures: mutable-only means you cannot reproduce a production incident; immutable-only means every keystroke is a commit and the editor is unusable.
Your tool has Git sync. It has no local branch, no merge, no pull-request review, and pulling overwrites unpushed local work. Someone loses a day.
The rest of this lesson unlocks when you commit above.