We're not talking about diversity of opinion on whether gravity is real. We're talking about viewpoint diversity on questions that involve far more ambiguity, complexity, and interpretation, like how to measure human intelligence. Questions where the evidence can point in multiple directions depending on what data is collected and how it is interpreted.
I think the site addresses this well:
> For simple problems or fully resolved technical matters there is little need for viewpoint diversity. Sometimes there is just one answer, or just one way to approach a problem. But for “wicked problems” — those that can be framed in multiple ways and that may trigger passions or partisan motivations–viewpoint diversity is essential. --https://heterodoxacademy.org/the-problem/
The problem with that concept is that if the available evidence actually does strongly support a position, seeking viewpoint diversity means actively selecting people who either cannot analyze the evidence or who wilfully ignore it's content.
Doing so within the research community may help the competent, and honest subset of the community avoid getting stuck in ruts formed by the subtle accumulation of confirmation bias from past results (if the other group doesn't end up getting filtered out by then over time, despite being present in the community, which is improbable), but it also means that you have to actively promote crackpots and their pet theories to do it; in terms of learning what is actually true, I don't think that's a net win.