Its preferred to fail the ductile member before the brittle connections because the failure is less catastrophic so you design the brittle to be stronger than the ductile.
That might be the plan to mitigate failure at the joint, but that does nothing to deny the fact that transitions and joints are a very frequent point of failure. The entire reason for over-engineering them, as you described, is precisely because of the propensity for those failures to occur there.
First off, now that I'm home and reread it, I didn't mean all this as dickish as it sounds.

None of the failures I've seen at joints have been caused by stress. Either shitty weld that cracked or fatigue.
