It is to some degree, isn't it? People don't personally replicate every study they cite and build off, they cite and build off studies that they believe to be trustworthy, part of which is determined by opinions of peers in the field (even if there's not a clear consensus one way or the other).

No. It's the opposite. All it takes is one single result to show a theory or body of "consensus" is in error.

No it does not. Science is full of contradictory results, that is why it is super easy to cherry pick studies.

You can't tell anything until it is in scientific consensus stage, precisely because a single study rarely means that much.

I'd argue that's more a characteristic of the publishing and peer review process in academia, not actual science.