For nearly any policy change, you can probably find 20% or more of people who would not have made that same exact choice. Many of them, if left to their own devices, will persist in disagreeing, slow-rolling, being mildly obstructionist, or similar.
If you don't have a clear (and enforced) principle that requires the implementation of made decisions and bars re-opening discussion absent material new information, you'll end up with an organization that lazily drifts towards the new decision instead of aggressively pivots towards it. That's clearly accepted in some companies; it doesn't seem to be accepted at Amazon.
That’s the whole point of disagree and commit. Most decisions are not clear cut obvious decisions, otherwise it’s unlikely there would be disagreements about them.
Disagree and commit allows everyone to contribute to their max capacity and move forward once deadlock has been reached.
There's only one place I can think of that might have benefited from that rule, and there were other solutions that would have worked as well if not better. Like a boss with a backbone.
Ultimately, if you need to exercise your management authority about stuff that your reports are doing that's a time-limited move, as too much of this will lead to reports just doing exactly what they're told, which is the kiss of death if you want to build good stuff quickly.
If you're dealing with obstructionist people, surely there are other ways to detect and address it than 'disagree and commit'.