[maxerickson]

If the current solution is unstable and relatively far away from other solutions, you shouldn't necessarily expect random perturbations to lead to a different solution.