It's not just about review though. It's at least as much about being able to get helpful results from bisects when things break, possibly on a remote machine.
To me, sort of the same reasoning apply: given that the intermediate changes do not make a lot of sense to me, having them doesn't really improve my ability to debug. If a bug appear in that group of changes, I'm more likely to debug it starting from the final version anyway.
In a sense reviewing and debugging are kind of a similar thing: you stare at the code watching for anything that might be off! :-P