I'd consider that to fall under my category 2

"changes to overall architecture or changes to the way the feature or code works large scale"

I'm not saying there's never a reason to go back and redesign something after a PR review, but in my mind getting a design to that point and then actually needing to change it is a huge failure.

Far more common the case where someone wants a big design change with no tangible benefits just different tradeoffs.

I just don't think the ocassional benefit is worth the cost of the process.