Say I work at a business that uses python exclusively. If I see my coworker get fired for suggesting we use haskell (or some other very different stack) for performance-critical code, how likely is it that I would later suggest we rewrite some stuff in go? or even python 3? or use https?
The opinion that biological differences between the sexes result in different career preferences isn't exactly wildly uncommon, extreme, or nonsensical.
I guess the question here is whether we think that the opinions in the memo are more like advocating Haskell or advocating ASP. They seem like the latter to me; if they seemed like the former, I'd agree with you. I don't think it's weird to think that there exist both rare defensible opinions and rare indefensible ones.
Note that the opinion in the memo isn't restricted to different career preferences correlated with gender (at least some of those opinions in the memo, like women wanting better work/life balance and men having rigid gender roles, are so uncontroversial that they're part of the standard feminist position too). The opinion also includes the claim of different abilities correlated with gender, because that's what's relevant to the business practices he's arguing in favor of changing, and in particular abilities relevant to qualification for engineering roles at Google. That's a much more extreme position.
The opinion that biological differences between the sexes result in different career preferences isn't exactly wildly uncommon, extreme, or nonsensical.