I looked at this again this morning... This isn't just the case of the wheel not having enough material by the hub. It's a low cycle fatigue crack. The loading is inducing an area of high stress concentration in a small area on the wheel (we are talking on the order of a mm or smaller), and the repeated load/unload cycle during racing is nucleating a crack.
For one, the edges of the wheel spokes are sharp without any radiusing. When you load to a point, your area available to absorb stress technically approaches 0, and the stress approaches infinity (while that seems physically impossible, it's the concept of stress equaling force/area... radii are necessary to distribute loads). It's been a while since I've messed with them, but most of the high end wheels like Fikse tend to chamfer or radius all their spokes.
I'm sure those wheels have been ran through FEA software and analyzed (a high end wheel company would be ludicrous not to). However, FEA is only as good as its operator, and if the mesh wasn't refined finely around the edges of the spokes, it won't pick up the concentrations.
Just my .02 worth from failures I've seen... I've done some failure analysis simulation on aluminum wheels for grad school projects.