In order to get a fast servo performance and a fast attenuation of disturbance effects, it is often desirable to have a high loop gain in control-systems. Unfortunately, this may contradict to the requirement on having a low loop-gain in order not to violate the actuator limits, which could cause windup problems. The trade-off between these two can to some extent be handled by means of conventional anti-windup techniques, but in general not over a sufficient range of operation. In order to reduce the problem further, it is proposed that the fastness of the closed-loop should be continuously retuned with respect to the status of the windup phenomenon. For this purpose, a certain 'nonlinear dynamic windup detector' is proposed, which uses the actuator in- and outputs for detecting the presence of windup, and produces a 'fastness variable' as its output. It is outlined how to achieve anti-windup by utilizing this fastness variable for on-line tuning of the controller. Typically this could be done by pole-shifting.