This has always been the case and this technology has been around for a while. I'm surprised Apple would have chosen to use it for user input.
one reviewer said they kept hitting "No" and "cancel" when they meant "Yes" because their natural progression was to look at the buttons in order, pick out the one they want to press but actually click it after they finish scanning all of them with their "mouse" paused on the one they want to hit. It's kind of fascinating that the VisionPro's input method works so well when it does work that it tricks people into doing this, because they really do just expect it to work like magic.
The problem is the Vision OS has no cursor. Cursor position is critical. Your eyes move around a lot more than your cursor moves, but in Vision OS, cursor position and eye position are the same. Very annoying. Maybe they could put a little cursor in the center of the screen that you could move around with another gesture.