You are grossly oversimplifying experimental design reducing it to the most uninteresting, idealized and trivial cases.
I don't care if you think I'm oversimplifying, or even if you say it, because unless you can show that you're correct, there's no reason to take comments such as this seriously. In point of fact, that is how experimental science actually works - the whole point of the scientific method is to simplify things as much as possible so that you're only testing one thing at once in any given experiment. That's what isolating the variable means, eh - you're isolating the thing you want to test and seeing how everything else reacts to changes in it.
How the actual fuck do you think an experimenter can isolate a variable on say the atmosphere, the universe or an ecosystem.... they can't and they don't, that's how.
I sure am glad that scientists don't have this attitude you're displaying of "can't do it, so it doesn't get done". What an experimenter does when faced with a tough problem like that is model it - whether it's creating an actual model or using a computer to do it instead. If you have a good enough model, you certainly can isolate the variable you want to look for. That's why science has gotten so far so fast, you know. The sort of empirical observations you're romanticizing are good enough to get a passing grade, most of the time...but sometimes a passing grade just isn't good enough.
Hunters, farmers, indigenous people bet their lives/livelihood constructing models based on empirical observation.....scientists do exactly the same thing on large complex dynamic systems.
Sure, they bet their lives, and often forfeit them due to the fact that the human brain is simply not very good at such things
. The survivors then rinse and repeat. We are very fortunate as a species that we reproduce fast enough to replace our losses, because those models they use are basically patchwork and guesses, held together with spit, baling wire, and duct tape. As I said before, this is usually enough to get a passing grade, but the real test comes when reality takes a sudden left turn and the model isn't ready for it.
That's where science comes in. Science isn't held back by the limitations of the human brain the same way as the empirical observations you romanticized, because those are generally limited to a single person or a relatively small group at best, whereas science is designed to be spread as widely as possible so it can use tools which simply aren't available to a single person.