For example, here is an article disputing it. Just one example of many.

Which all do recognize some general upper limit, even if the variance can dip into the single digits or approach the high triples. The point being that there is a functional upper bound, and certainly not one so high that it can accommodate a fully high school's worth of students much less a nation's worth of citizens.

No, I don’t think reducing the tolerance paradox to biological limits is productive or instructive.

Its useful from a practical perspective, as it demonstrates a real upper limit on the individual. For the same reason that estimated life expectancy, standard walking speeds, and normal sleep patterns shape our basic expectations of human behavior and comfort, an understanding of social maximal empathy limits can help us engineer social structures efficiently.

You wouldn't expect a normal human to sprint at the speed of freeway traffic. Why would you expect a normal human to empathize with a constituent group of a million people?

We see this kind of rhetoric in all kinds of extremism

We don't just see it in extremist ideologies. We see it in every ideology. Milquetoast moderates like George Bush and Bill Clinton had the same fundamental impulses when they governed the US as Ralph Nader and Ross Perot. Only their policies differed. Policies that were inevitably most favorable to very particular constituencies. This was not a difference in their scale of empathy.

Nativism and alienation will always be a problem for groups of humans at the scale of thousands. And so social and political structures need to be resized to accommodate that upper bound. Otherwise, tolerance just becomes double-speak, a term you toss about when you're angry at some out-group for failing to conform to the biases of your in-group.