Well Actually to do the last thing - a naive approach is much better than tackling with C++ and Precision:
I found my self having 2 Audio Components (and using them as a double buffer) - each time a Time Dilation changes I am fading out current AudioComponent set correponding Sound to another sound component and fading it in - and the Precision is good enough!