We hit 86–88% on bench, but at 200 L in a jacketed glass-lined reactor we stall at 78% and see impurity B spike when bumping agitator from 180 to 280 rpm and using a 2 C/h ramp. Has anyone recovered selectivity by switching a 10-minute neat charge to a 45-minute semi-batch feed or by moving from three to four baffles to tame vortexing, and do you target constant tip speed or power per unit volume during scale-up?