I get why this pattern catches people's attention, it really does look suspicious when you line up the cases. But I think we have to be careful here because confirmation bias is a real thing - we notice the suspicious timing and remember it, and we forget the hundreds of witnesses who testified just fine.
That said, I'm not saying dismiss it entirely. Some of these cases genuinely do have weird details that official investigations kind of glossed over. Epstein is the obvious one everyone brings up and honestly the security camera "malfunctions" alone would raise eyebrows in any normal context.
What I'd want to see is someone doing a proper statistical analysis - like what's the actual rate of witness deaths vs the baseline population. Has anyone seen anything like that done rigorously? Because right now its mostly pattern matching and I think the argument would be a lot stronger with actual numbers behind it.