"How does this pill work?" -> "You put it in your mouth" comes from observing at the level of your mouth, or alternately observing the intake process. It's not the most insightful detail to observe about the pill, but it's probably the most practically useful!
You could of course put the pill somewhere else, and maybe it would work there too or maybe not (it probably depends on where exactly you put it).
This bugged me, but I would get worn down by it and sort of give up the "search for truth". I feel bad, and then I rest, or take ibuprofen, or drink fluids, and my body just fixes it eventually and the mysteries continue.
Doctors don't make your body better. They only make your body less worse than normal. And those sound like they should be logically the same, but they aren't.
Indeed, science provides scant information about how drugs work and that includes medications with effects (favorable or not) in the brain. The reality is that drugs have a great number of effects very few of which receive adequate study. IOW while there are useful target effects that are exploited, the fact that drugs are more or less tolerable is largely a matter of luck.
This reality is illustrated by instances of tragic drug outcomes despite long use in clinical practice. Of course observing for and identifying random events (adverse effects) depends on the sample size. Registration trials generally use hundreds to a few thousand people which is highly unlikely to reveal serious, rare outcomes occurring on the order of 1/100K recipients. Monitoring of after-market AEs is necessary to detect such anomalous outcomes.
The upshot is that clinical practice is at least as much art form as reliant on science. But as you say this is also the case with all science, albeit to a variable degree. It implies your statement should read "science can't answer how anything actually works" since there's no way to know what is not yet known.
That doesn't mean we have no useful knowledge, far from it, as the vast array of extant productive technologies clearly shows. But in truth we can't fully "explain" anything, we have to be content developing usable "working solutions" to real-world problems. To my mind process of science is satisfactory to the extent we accept the inherent constraints of human knowledge. With that acceptance we gain the flexibility to move forward, to find new and better working solutions, and add insightful observations to the knowledge base of science.
I'd argue that this is not true. Science deals completely and exclusively with modeling "if this then that". We don't know how gravity works, but we know that it does appear to work, and we believe that it works in the same way every time, and we can use observations and organized thinking to model gravity in a way that allows us to predict what will happen "if this".
And the breakdown for "if this then that" depends on the current observability of the scale you're trying to describe. We can safely say we don't know "how" antipsychotics work inside the brain, but we can also safely say we know "how" antipsychotics work on populations, because populations are easily observed and inside the brain is not.
This is normal.
At an essential level, science can't answer how many things work, because, the deeper you go into the unobserved (and the workings of the brain are still approximately 100% unobserved), "how" becomes more and more a mis-use of the word "how" to talk about "why" something works, and science can't answer "why".