Why not both? Most research directions ... are useless, don't pan out, don't have the tools to actually achieve anything yet, won't ever get finance for necess

Why not both? Most research directions ... are useless, don't pan out, don't have the tools to actually achieve anything yet, won't ever get finance for necessary tooling, ... and fail. Useless is the least bad of these. Nevertheless, papers on these are important for a lot of reasons. And one in hundred research directions succeeds. And then there's tons of "refinement" papers. First reason: you won't know which is which until you try. Second: researchers learn a lot even when exploring things that won't work (in fact this is often done as exercise). Third: these allow institutions, infrastructure, ... to exist, to be ready when there is something that does pan out.