Ah. I'm not super familiar with that "bug" - is it that "touchmove" events don't get fired for single-finger scrolling (panning)?
Perhaps you could make it work (albeit less gracefully) in iOS by reconciling the header visibility state once "touchend" finally gets fired though? Assuming "touchstart" gets fired as you'd expect, you can do a bit of manual calculation to determine if the y-position changed enough to warrant header hide/show.