You make multiple assumptions here
1) You assume that the during the time that passes between the LED breaking and the user noticing, there was not a single attack or a single blunder that caused the camera to turn on and record/capture something that was unintended.
2) You assume that the LED breaks deterministically. The LED can break randomly. Maybe it lights up when nothing is being recorded resulting in a false positive. The user has no way of differentiating between a false positive and a true positive which can result in unintended captures.
3) Similarly the LED can break in a way where it sometimes doesn't light up when something is being recorded even though power is always sent to the LED when the camera is on resulting in a false negative. Again, the user has no idea of differentiating between a false negative and a true negative.