Hm, the first step is fully enforce Open Science, and, when relevant, pre-registration. This would make fraud much harder to begin with (though not impossible of course). Then, those "detectives" would (hopefully) have a manageable workload.
Pre-registration would be great! There is some case where there is a collaboration with the industry where the goal for them is to patent and then publish. I’m this case a public pre-registration would be problematic. How would do handle that case?
Sorry, I couldn't really help you there, I'm working in evolutionary biology, we don't do patents and industry much. However, I would tell that I don't believe pre-registration is the silver bullet for any research. Even beyond your case, there are much research in my area that is not hypothesis driven, more exploratory or completely inferential (which is totally OK!), for which pre-registration makes very little sense.
Academia