That's a good and subtle question, with multiple relevant answers: 1) As _Microft says in a sibling reply, the reason is because for this type of experiment, t

That's a good and subtle question, with multiple relevant answers: 1) As _Microft says in a sibling reply, the reason is because for this type of experiment, the researchers were studying short-baseline oscillations. A curious property of neutrinos is that they oscillate -- we interact with them through their "flavor", but they propagate through space in a mixture of flavor eigenstates, so a neutrino that is created as an electron-type neutrino can interact later as a muon-type (or tau-type) neutrino. LSND and MiniBooNE study short-baseline oscillations, where L/E (the distance travelled divided by the particle energy) is very small, a few meters on a ~1MeV neutrino. By comparison, almost all the (vast number!) neutrinos that pass through each of us each second come from much farther away. Most of them come from the Sun, so their origin is distributed over a sphere 100x the diameter of the Earth (and, to make matters worse, something called the MSW-effect scrambles things further). So, people who study these effects either build an accelerator or snuggle up next to a reactor. In principle, one could use the existing neutrinos from the Sun to do this sort of experiment, but one would need a material capable of blocking all of, say, the electron-type neutrinos from the Sun. If such a thing existed, we could place a detector a few meters away from that filter and watch for electron-type neutrinos to reappear. Alas, the only known material perhaps capable of achieving such a task exists for a fraction of a second inside core-collapse supernovae -- the extremely dense infalling matter is so dense that it can actually trap some of the outgoing neutrinos.... until the neutrino pressure blows it apart. 2) In this specific case, the experiments were originally done with muon anti-neutrinos. The easiest way to get those is with a particle accelerator. 3) One can also attempt to get at this sort of thing (though the approach has different sensitivity/systematics) through reactor-neutrino experiments, as one of the downvoted posts below suggests. For the shortest baselines, though, there's probably really nothing like a beam-experiment. 4) (an aside) One thing that Hossenfelder doesn't really address in the article: There are still systematic-uncertainty questions around LSND and Mini/MicroBooNE. The modern experiment has addressed many of them, but not all. The experiments are difficult and the anomalies have their quirks. Yes, the combined result reaches a statistical power of ~6-sigma, but I don't think you'll have to look too far to find credible experimentalists who have concerns about the result. This is in contrast to something like the Higgs, where the discovery was fairly clean and simultaneously confirmed by two groups. LSND was huge news back in the day, but the difficulty/expense of replication and the substantial experimental challenges mean that the situation remains fairly murky.