The only thing I can think of is that the cases are separating at the juncture of where annealing stops and stiffer brass starts -- i.e., a materials/stress transition point.
Normally I'd expect the effect to be graduated down the case body. But maybe not. (I could envision induction heating as producing such a sharp transition).
Is there a possibility of your annealing a dozen cases/extended shoulders "the traditional way", and then tracking their failure rate compared to the others?