As reported by pspeed, pr #875 added a nullcheck that changed the behavior of positional AudioNodes that aren't attached to any parent.
While this is not the most common way to use audio node, i think it would probably be better to restore the previous behavior, since it doesn't affect the topic of the pr.