ReplicaSetAwareServiceRegistry executes various hooks in the order they are registered. There might be implicit dependencies between those hooks that are not surfaced just because they happen to be executed always in the "right" order. We could randomize the execution order of those hooks, at least just during testing.

The type of issues we'd be interested in catching would be things like SERVER-89285, that initially introduced a change that brought the catalog into a consistent state at that point, but some other hook might be already relying on the consistency of the catalog.