gonzalo75: Another possible issue is the inefficient shape of the stiffener. The current concave profile of the overstressed stiffener causes material to be missing, where material is needed. If you instead use a straight (not concave) stiffener profile, with only a slight (not rapid) depth taper, then the stiffener might be much more efficient.
What aluminum alloy (and temper) is this? Unless I am misinterpreting, a 2 % rupture strain sounds brittle, in which case stresses greatly exceeding the yield strength might not redistribute well.
You say it would take no money to make a change now, but you said the analysis is very expensive. Therefore, unless I am misinterpreting, it sounds like your posted stress analysis is not one of the expensive, outsourced, final analyses, right?