>You can directly weed out poor performers because you can watch them perform poorly.

Why don't we fire researchers who committed fraud or who are incompetent?

Some studies will fail replication, even though the researchers did everything right. But in many cases the methodology is flawed, statistics are misused or the data is massages. In industry this would be grounds of termination and this isn't even the outright fraud, which also happens. For most of these no serious punishment is expected.