I agree to this (and not only because he's the Old Man) that perhaps the cut out modelling left out inertial behaviour. Ingame it had formerly cut out instantly when the correspondig g was achieved while according to my understanding of the Hurri pilot's quote it was more inert and took some time to take effect.

So if I were dev I would have tried the following:
keep the historic g values for the different modes of cut outs. Apply the corresponding effect in full extent (100%) on engine performance with a time delay of 2secs as described in another thread to be the experienced delay in RL. Apply a smooth transition of mixture leaning between the moment of appearing g level and the full effect of mixture ratio. But this smooth transition should be not linear but rather parabolic or exponential with time. For simplicity one could approach this non linear behaviour by segmenting the transition curve let's say into three of four segments and inbetween the segments it is linear. This allows small gradients in the initial stage and steep gradients when approaching 100% effect of the corresponding cut out level. Then test if it is still sensitive to turbulences and small trim tab movements.