A little of both. The problem is that evaluation of research work is insanely difficult. A lot of people think it's easy because "the world is objective, it either works or doesn't" but research is cutting edge and you're only chipping away at a much larger picture. It can take decades for a work to reveal itself as truly profound or utter shit. The problem, which I rant about in a longer comment, is that instead of acknowledging the noise we've embraced poor metrics and encouraged the hacking of those metrics. I call this Goodhart's Hell. People forget, metrics are models and all models are wrong. You have to constantly be questioning your metrics and determine how well aligned they are with your goals or else you'll drift (the environment moves, so your metric must move too).

I think actually the better way to solve this, which may seem paradoxical, is to actually increase funding. Not in size of single prizes for grants (well... we need that too, but that's another discussion), but in the availability. The reason being that the hacking is partially encouraged by the competition for a very scarce resource. A resource that compounds. Due to this (and some nuances, see other post) we're not actually rewarding those who perform the best work (we may actually be discouraging that) but those who become lucky. A "good work" is simply one with high citation counts, which is heavily weighted on the publicity around that work. Which is why top universities have big media departments, pay news publishers to advertise their works, and why survey papers generate huge counts.

The problem is that the system is rather complex and there are no simple or "obvious" solutions. "Good enough" is also not clear because too low order of an approximation can actually take you away from your intended goals, not a small step towards as one might think.