Well, this is obviously not true. As we change dmeshutter right before the frame and always get it correct with the new shutter value. As for dmegain it indeed requires to be set 1 frame in advance (not 2). Although I do not understand why it works this way if we are setting those values directly to the sensor and what is the reason of making direct settings to the sensor as opposed to setting it in the ISP if it would converge only after N frames even with direct setting.
Anyway the problem we are experiencing is with lost frames rather than incorrect shutter/gain values.