One downside of Auto-Ducking is that you can't fade back to normal volume before the sound has finished. The only signals it can read is Start and End of a Sound Source. But if you could optionally set Markers in Sample-Editor or even perhaps set the Loop Start / Loop End Region in Sample Editor as readable signals, that would make Auto-Ducking a lot smoother. I can't imagine that causing significant performance impact, yet it would make it sound a lot more accurate to the desired result in so many cases.