It doesn't matter. When it comes to the fantasy RPG, Dungeons & Dragons, I will fight for the normalization of any genders and playable species to be able to have any ability score that can be achieved by the same ability score generation methodology in the core rules. Science (or "realism") be damned. There can be any fantasy explanation for these results that transcends the limits of our world's biological science (we only have humans in our world). And I think Wizards leans towards that way of thinking.
"Real world biology" is an argument I wouldn't even consider. If someone wants to limit female or halfling strength, or orc intelligence, they can force those exclusionary preferences into the microcosm of their own house rules or games. But it doesn't belong in the D&D core rules.
But, uh, that's not what's being argued here.
The fact of men being stronger than women (both in average and at the extreme) does not require that a game incorporate that. But the game not incorporating that does not mean that the original fact is false. There are certain assertions being made that the original fact is of questionable veracity (which is a lie), partly because they do not want that fact to be used to justify the inclusion of that difference in the game (which is a game design issue, not a "bad science" issue). However that's an (incorrect) assertion about reality based on what you want in the game, not the other way around.
~~
Every game incorporates realism, even down to something trivial like Chutes and Ladders — you're climbing the ladders to go up, and sliding down the chutes to go down, because that's what makes sense to minds that are used to the real world. Every game also limits how much realism is represented in the game. Chutes and Ladders does not try to model the exhaustion you'd suffer from climbing all those ladders, for example.
An RPG spends a lot more effort trying to model reality. (In D&D's case, that starts with the six basic stats.) Thus the question of whether men and women should have different default stats (or caps, or whatever) is perfectly reasonable. The answer, however, has nothing to do with reality, and everything to do with game design. Is this a modeling of reality that helps make the game better, more enjoyable, or more useful mechanically? (Note: "Is this more realistic?" may help in choosing between options, but is not itself a primary game design question. It's a modeling question.)
People who like creating models like putting numbers on things. Orcs are stronger, elves live longer, gnomes can't walk as fast, etc. When first building such a model, having different strength scores for men and women is not exactly controversial. Some games also have degrading stats as you age, or lower stats available for very young characters.
However the purpose of an RPG is not to parade around your model of reality. You want to simplify that model as much as possible, in the service of the
game and the people playing it. You want to address pain points and polish rough edges.
Can a dead average 10 Str person carry around 150 lbs of gear all day without being slowed down? Of course not. But a more realistic model would be both more complicated and less useful for the enjoyment of the game, so a less stringent adherence to reality is used. Likewise, are different strengths for men and women
useful to the game? If no, or at least no for the vast majority of cases and players (and especially if there are benefits to ignoring the distinction), then it shouldn't be included.
The same sort of argument has led (or at least could lead; actual motivations are fuzzy) to changes in how racial stats are allowed or assigned. Tasha's makes them freely assignable, while Level Up shifts them to the character's background. Both have a similar effect, and are arguably more realistic than the original, so an improvement of the model.
It's a modeling vs gaming issue, and the model needs to serve the game, not the other way around.
~~
Aside: I'm a computer programmer, and I build models of stuff
all the time, so the above seems blatantly obvious to me. Every time I write a program, I'm essentially going through the same process: create a model to match the problem (or at least my understanding of it, which may depend on other people's understanding of it); figure out what isn't working right and adjust it; get rid of unnecessary noise (non-useful bits of the model); and optimize to make everything run as smoothly as possible.