If "male visuospatial advantage" isn't being replicated when stereotype threat is taken into account, and stereotype threat isn't being replicated when other experimental variables are taken into account, I don't think that necessarily proves or disproves male visuospatial advantage or that it can or can't be explained by stereotype threat.
A meta-analysis wouldn't solve the problem unless it somehow included previously unpublished data. Publication bias has left whole fields of science untrustable.
If an individual scientist cherry-picked interesting data the way publications cherry-pick results to publish, it would be called fraud.
I think we need some meta-analyses here.