The best design for what you describe (in my humble opinion), is the following:
A multi-instrument with the main sound, async and loop buttons checked, with a sustain point snapped right at the end of the instrument, and either a finishing sound right after that, or an AHDSR modulation on the main instrument volume (to set the release). Instead of stopping the event, you do a “key off” (and the event will dye by itself).
I say multi-instrument because single-instrument seems to have a bug (it cuts even without the cut button checked). But it also could be any other instrument type (I use a scattered instrument for a footsteps loop, for instance).