Using DAMON for proactive reclaim

IMHO, these days, any time you have a control knob, that's just a learning algorithm screaming to be born. We should be using closed-loop feedback-based control algorithms --- or just machine learning --- to tweak knobs, not slow human meat brains.