AFAICT meta-studies is the level where we as a society really can try to say something intelligent about how stuff works. If an important question is not included in a meta-study, we (i.e. universities and research labs) probably need to do more research on that topic before we really can say that much about it.

But publication is not a closed system. The "published, peer-reviewed paper" is frequently an artifact used to decide practical policy matters in many institutions both public and private. To the extent that Science (as an institution in its own right) wants to influence policy, that influence needs to be grounded in reproducible results.

Also, I would not be surprised if stronger emphasis on reproducibility improved the quality of conversation among scientists.