Steps to reproduce: 1. Visit https://www.scirra.com/labs/bugs/pointer-frames/frame.html 2. Place a pointer down (using mouse on desktop or touch on mobile) inside the frame on the left, drag to the right outside the frame, and release 3. Observe the events printed in the frame The frame should see either a pointerup event when releasing, or a pointercancel event when it loses the pointer. Both Firefox and EdgeHTML fire a pointerup event (but Chrome has the same bug). Instead, the frame receives neither a pointerup or pointercancel event. In more complex web apps, this causes code that tracks pointers to get permanently stuck, since it will never see an event indicating the pointer ended. This only happens when preventDefault() is called on the pointerdown event. However sometimes this has to be done for unrelated reasons. It should not cause a framed page to lose pointer events.
Note this also reproduces in Safari TP 97.
<rdar://problem/58119021>