No, theoretical research is needed to understand why a gizmo works, we often use phenomenological models to progress, and the fundamental reason that ties everything together is figured out later. We didn't understand electricity and magnetism for a long time but we were creating all kinds of devices and materials via trial and error. The arrow of discovery has a strong record with experiment first, theory later. Metallurgy was pretty much figured out by trying shit out first, later came the formalization and explanations for why strain hardening is a thing, or doping and microstructure for example.

For the humanities, yeah I have no idea. I didn't specify but I meant the science and engineering parts of academia. I think math doesn't have a reproducibility crisis since it should be all logical proofs that don't require infrastructure, so I guess math is fine as it is.