@TimCampion to offer a different point of view, i would regard this as a defining property. namely, one of the fundamental problems that quasicategories address is that homotopy equivalences aren't invertible in the strict sense. the two facts that levelwise equivalences do become invertible in the coherent sense, & that coherent diagrams can be made equivalent to strict diagrams, makes coherent category theory immediately interesting & useful from a classical point of view
so if we didn't have that property, we're back to zigzags of diagrams, or calculus of fractions, or model categories, or some other technical tool to address it in a different way.
I think what I find confusing is the following. In $Fun(BG,Spaces)$, every object is equivalent to one where the $G$-action is free. But in $Fun(BG, Set)$, this is not the case -- even though $Set$ is a full subcategory of $Spaces$.
I suppose this is because I'm equivocating on what "free" means. But it still feels weird.
@TimCampion I don't know what you mean by a "free" action. Not every object is equivalent to one such that the homotopy fixed points are empty, for example
In $\mathrm{Fun}(BG,\mathrm{Space})$, $EG$ is actually equivalent to the point with the trivial action, so there's not really a sense in which it is "free" (unless by "free" you mean cofibrant, in which case it becomes obvious that's a model-dependent notion)
@CharlesRezk Wow! Is Vandiver's conjecture the same as the Kummer-Vandiver conjecture? The one equivalent to calculating $K_{4n}(\mathbb Z)$?
@DenisNardin Yeah, I'm not sure what I mean exactly. But it feels weird that in higher category theory we have this distinction between genuine vs Borel equivariance whereas I think the two coincide in ordinary category theory?
@TimCampion I mean, it is a distinction about what maps you invert in your localization. Yeah, it is interesting, but I don't think it's useful to see it as a distinction between 1-categories and ∞-categories
Maybe that's fair. If I take the 1-category of $G$-spaces and take generating cofibrations to free $G$-cells with the usual interval object, I'll get a model structure for Borel $G$-spaces where the equivalences between cofibrant objects are $G$-homotopy equivalences...
I guess I'm thinking that $G$-homotopy equivalences are the "obvious notion of equivalence"
So from that perspective it comes down to what you think the obvious cofibrations are...