seems like I've heard that faults are overdue hundreds of times, but never seen a demonstration that overdue faults are more dangerous than underdue faults. What am I missing here? Are the past earthquakes so much like clockwork that overdue in this particular case means more dangerous?

Admittedly, if the recurrence interval is shorter than previously thought, the odds of a quake go up, but then the expected magnitude then goes down, as you note. And if less slip is expected, then by the simplest thinking (Wells-Coppersmith relation), the rupture on average would be shorter than previously thought.

And the idea that we now have a firm grasp of what does and doesn't stop a rupture seems optimistic.

It is an interesting idea, but from what I know, is more research than anything to raise an official alarm about.