BTW I would expect that such a test would say

QCOMPARE(textEdit->toPlainText(), QStringLiteral("New title"));

rather than using model.property() without really knowing what's going to come out of it.
(e.g. if the change didn't happen in the model, this would still pass).

But I guess you consider the stub model to work, as it's part of this unittest, i.e. not the thing we're actually testing here. Still, a wrong value for shouldReactToNotifications and this would still pass, with a wrong expected value...

In general I prefer expected values to be constants than values returned by methods, for this reason.

That's indeed correct. I think we did that at several places, we kind of assume the stub is "perfect" to avoid repeating ourselves on the data. Still that's somewhat dangerous we should probably use const variables used both in the GIVEN and the THEN.

Hmm? This looks like a unittest for artifacteditormodel to me.