The incapability of thermal models to accurately reproduce the horn-like structure of the Kaon-to-pion ratio measured at AGS, SPS, and low RHIC energies, as well as confirmed in the beam energy scan program, has long been a persistent problem. This issue is believed to have arisen due to the inappropriate application of statistics, particularly the extensive additive Boltzmann–Gibbs (BG) statistics. The assumption that the analysis of particle production, a dynamic non-equilibrium process, should be primarily conducted using extensive BG or non-extensive Tsallis statistics, has proven to be an unsuccessful approach that has been followed for several decades. By employing generic (non)extensive statistics, two equivalence classes [Formula: see text] emerge, thereby undermining the validity of any ad hoc assumption. Consequently, the degree of (non)extensivity exhibited by the statistical ensemble is determined by its own characteristics. This encompasses both extensive BG statistics, characterized by [Formula: see text], and non-extensive Tsallis statistics, characterized by [Formula: see text]. The energy dependence of light-, [Formula: see text], and strange-quark occupation factor, [Formula: see text], suggests that the produced particles are most appropriately described as a non-equilibrium ensemble. This is evidenced by a remarkable non-monotonic behavior observed in the [Formula: see text] horn, for instance. On the other hand, the resulting equivalence classes [Formula: see text] are associated with a generic non-extensivity related to extended exponential and Lambert-[Formula: see text] exponentially generating distribution function, which evidently arise from free, short- and long-range correlations. The incorporation of generic non-extensive statistics into the hadron resonance gas model yields an impressive ability to rightfully reproduce the non-monotonic [Formula: see text] ratio.