For 1 and 3 I think that the question is do these examples (assuming that they are real and true) support the argument made, and can we , do we, agree that it is plausible and reasonable to assume that these are real and true. If you doubt the notion that a social worker could be dismissed by this kind of data driven change of policy then I would concede that damages the argument. This is not a physics observation though (kinda the point here!) if you say that there is one example where a social worker wasn't dismissed in this way then I do not agree that is sufficient to derail the argument. On the other hand if you show me one example of a situation where an atom decays differently and produces behaviour not accounted for by the standard model I will happily declare that the standard model is bust.

For 2; this is not about the contingent knowledge that science produces, it's about the applicability of the process that produces that knowledge so effectively in some domains. There are domains where it is not appropriate or useful to think exclusively and uncritically in a scientific way, medicine and economics being just two examples of domains in which uncritical application of the scientific method and mindset have done significant harm to a significant number of people.