I don't follow. An experiment needs to be repeatable in principle. You don't actually have to re-run it with the exact same specimens or subjects; they can be

I don't follow. An experiment needs to be repeatable in principle. You don't actually have to re-run it with the exact same specimens or subjects; they can be sampled from some larger population that you're studying (e.g., patients with depression). Nevertheless, you can often do something like a cross-over design that lets you repeatedly test the same patients in different conditions.