My post assumes that because there's no good evidence that, as you put it, "reality is sexist" in the relevant ways and that we should reject the null hypothesis. There are plenty of plausible alternative hypotheses (the Damore memo vaguely alludes to them), but I have not seen strong reasons to accept them, just post-hoc rationalizations like "the average woman scores worse/better on $metric, so here's a story for why the job requires more/less $metric". Tellingly, those rationalizations have changed as programming went from a low-status to high-status position: there was folklore several decades back about how programming was obviously women's work because it was like dinner planning.

If you do find a scientific reason to reject the null hypothesis, hopefully such an analysis will come with some specific number other than 50/50 - and we can see if Google's processes match those numbers.