Simplify "strong limiting" logic in limiter

While writing the limiter specification, I noticed that the logic used to trigger strong limiting can be simplified without affecting BE. The resulting code is easier to understand and to describe in the spec.