Why should you test an experiment only one variable at a time?

It is a matter of certainty. If you change only one variable and the outcome differs, then you may safely assume that the change in the one variable was responsible for that change in outcome. If you change more than one, then how would you know what was responsible?

You wouldn't. You would be left guessing. One of the objectives of good science is reduce the guesswork down to as close to zero as possible.