In classical mechanics, we often wish to calculate a partition function of, say, an $N$-particle system $$Z = \int_\Omega {\mathrm e}^{- \frac{ E(\vec x) }{ k_\text{B} T} } \, \mathrm d^{6N} \vec x .$$ This integral is taken over a $6N$-dimensional phase space $\Omega$, with elements of type $\vec x = (\vec q_1, \dots, \vec q_N, \vec p_1, \dots, \vec p_N)$.
Because particles are indistinguishable, the energy depends on $\vec x$ in such a way that permuting any two position components (as well as, simultaneously, the corresponding momentum components) changes nothing.
I have the feeling, however, that you could bake this information directly into the topology of the phase space you integrate over, so that two points $\vec x_1, \vec x_2$ with particles permuted land in the same point of phase space. What would this space be, topologically?
I apologise in advance if this question already has an answer; I couldn't find it. This answer discusses the topology of phase space given Hamiltonian constraints, but this is not what I am looking for, as it doesn't talk about the Gibbs correction. This talks about the Gibbs indistinguishability factor being an approximation, but no mention of the topological perspective.