Could it be possible that using a ball as a cursor could result in the cursor moving so quickly that it leaves the ball's box-model before the mousemove event has a chance to trigger again?
Indeed, that is a possibility.
This situation is more likely to occur when dragging an object upwards from the top row of pixels of the object.
However, this problem is less probable when dragging a larger object from its center with a computer equipped with a high poll-rate mouse, high display refresh rate, and a browser capable of processing OS-provided input events rapidly. For instance, front-end developers using a high-end gaming machine with a 1000Hz USB mouse and a 120Hz+ display may not experience this issue compared to those using a 100Hz mouse on a 30Hz display (which is quite common among users with 4K displays running on HDMI 1.x at 30Hz).
Another scenario where this issue might arise is when a user is using an absolute pointing device like a touch-screen or graphics tablet in absolute-mode (also known as "pen mode" for Wacom), rather than in relative-mode (known as "mouse mode" for Wacom). If a user simply taps the destination location with their finger or moves the stylus vertically outside the detection zone to "drag" an object to another screen location, only one mousemove event will occur and it won't be captured by the 'ball' at all.