Great point. It is definitely much harder to change many variables than one.

I think that what is the correct thing to do depends lot on context and importance.

For example, picking the right government system, or preventing rape, are worth a lot of effort in my book. And worth some seriously challenging systemic changes. In other cases, it may be best to only consider what can be accomplished with single variable shifts, and to throw out more complex possibilities.

Also, I think a lot of things you simply can't shift by only changing one thing at a time, so in those cases, it is many or nothing.

One thing that helps is to do a lot of study before making changes. If you can run models ahead of time, that is awesome. Then the cost of looking at many variables is greatly reduced, because its not a big deal if a model fails.