That makes sense binding to the smallest viable control surface, and the sampling strategy for hot paths sounds like a pragmatic balance between latency and coverage. Thanks for the additional feedback here.