While it is possible to use a scatterer instrument for this, the issue is that the individual sounds spawned by a scatterer instrument are never untriggered. This means that behaviour that depends on being untriggered, such as the release period of an AHDSR modulator, will never occur - and so your playlist entries will be able to fade in, but will always end naturally rather than fading out. If you use a scatterer, the only workaround for this issue is to build the fadeout into each of the individual playlist entries, either by creating custom timeline automation curves within a referenced event or by baking the fadeout into each of your source audio files, and either option would require you to manually design the fadeout for each and every individual playlist entry.
Using a multi instrument avoids this limitation. You could still get irregular intervals between sounds by using multiple loop regions of different lengths, and assigning all but the longest of them a probability trigger condition.