Modern astronomy, for example, can pretty much as well challenge the standard of randomized controlled trials: no one uses experimental planets and galaxies to test their null hypotheses.
Modern astronomy and astrophysics is just about the most rigorous experimental science outside of particle physics. Models are developed against simulations and past observations. Then new observations are proposed, selected, scheduled, and performed. The null hypothesis is almost always based on the standard models and can only be overturned by new models using new data.
A future observation of some phenomenon "out there" is, in principle, no different from a future observation of some phenomenon in the lab. We don't call them "experiments" but they are every bit as difficult to falsify. Perhaps even moreso, since those who collect the data are generally not the same people as those who design and test the models. Since data is eventually released publicly, anyone is free to re-run the simulations and re-test the models against the same data, as well as propose future planned observations to test any weaknesses in the models.
Modern astronomy and astrophysics is just about the most rigorous experimental science outside of particle physics. Models are developed against simulations and past observations. Then new observations are proposed, selected, scheduled, and performed. The null hypothesis is almost always based on the standard models and can only be overturned by new models using new data.
A future observation of some phenomenon "out there" is, in principle, no different from a future observation of some phenomenon in the lab. We don't call them "experiments" but they are every bit as difficult to falsify. Perhaps even moreso, since those who collect the data are generally not the same people as those who design and test the models. Since data is eventually released publicly, anyone is free to re-run the simulations and re-test the models against the same data, as well as propose future planned observations to test any weaknesses in the models.