Ian, you make an unwarranted claim (through a logical jump) that herding results in a single herd. Both empirical evidence and logic suggest otherwise: several competing herds may (and often do) emerge. Moreover, there is often an advantage from moving from a larger herd to a smaller one, both from a fingerpointing and a security (direct risks) point of view.
Certainly, there is no incentive for the smaller herd to give up its distinctiveness: if security breaches happen within the large herd, the members of the small one can exclude themselves (to some extent) from paying the extraordinary costs. Also, rational attackers can be expected to invest more in searching for vulnerabilities in the largest herd, thus stabilizing the smaller ones.
In short, the (otherwise very convincingly described) process of herding does not imply that only a single herd is stable. Nor does this seem to be the case.