I think the axis of difference here between the successes and failures is something like scaling and replicating-with-variations of things that have worked, vs scaling and duplicating things you think will work (and in many cases, are then committing to claiming did work and not being willing to update quickly). The latter case is much more prone to failure at scale from complications the planners didn’t model well. The former usually lets ideas that don’t work die before they get big enough to cause too much tragedy.
I think the axis of difference here between the successes and failures is something like scaling and replicating-with-variations of things that have worked, vs scaling and duplicating things you think will work (and in many cases, are then committing to claiming did work and not being willing to update quickly). The latter case is much more prone to failure at scale from complications the planners didn’t model well. The former usually lets ideas that don’t work die before they get big enough to cause too much tragedy.