Some people’s identifications are more reliably correct in some taxa, so it seems like an easy shortcut to check how reliable the person seems and choose to agree or not agree based on that. But nobody is 100% reliable in any taxa. If you instead check the direct evidence (the photos, location, date/time, etc.) and arrive at an identification based solely on that, then the reliability (or unreliability) of the previous identifiers won’t, and can’t, have any effect on your answer. That kind of independent identification is way, way more valuable than quickly clicking “Agree”, because it can be used to correct errors. It’s kind of like doing a long math homework question, writing down the answer, then throwing out your previous work and redoing the whole thing from the beginning. If you arrive at different answers, then you know you must have made a mistake somewhere, and if you arrive at the same answer then you can be more confident that you did not make a mistake. But if the person at the desk next to you thinks to himself “KitKestrel is reliable, so I’ll just copy down her answer” then having two identical answers cannot increase your confidence that the answer is correct.
So that’s what point 5 is hinting at (not strongly enough, in my opinion). In an ideal world, getting to Research Grade would require a certain number of independent IDs, but since there’s no easy way to tell the difference between independent and dependent IDs, all we can do is encourage people to do the work instead of taking the shortcut. Honestly, I hope I never get the reputation of being an expert in a taxon if it means I’ll stop getting independent verification of my IDs. I and all the reliable identifiers I know have made some mistakes that seem pretty boneheaded in retrospect. We wouldn’t have figured out they were mistakes if we hadn’t double-checked each-other’s answers.