The high-temperature, strongly nonlinear hysteresis behavior and underlying mechanisms of Entangled metal polymer rubber (EMPR) have not yet been systematically elucidated, posing a bottleneck to its engineering applications. To investigate the mesoscopic-scale high-temperature tribological mechanisms of EMPR, wide-temperature-range self-contact friction tests were conducted on 304 stainless-steel wires. The results show that, under the combined effects of material softening and high-temperature oxidation, the coefficient of friction (COF) increased from 0.3519 at 25°C to 0.4918 at 500°C. Based on virtual fabrication, an EMPR model with a relative density of 0.25 was constructed, and variable-COF quasi-static compression as well as thermo–mechanical coupled finite element simulations and experiments were performed. The roles of nonlinear elastic restoring force, dry friction force, and hysteresis memory force in the energy storage, energy dissipation, and delay mechanisms of hysteresis behavior were clarified. The quantitative mapping relationship shows that, under temperature elevation, the energy dissipation ΔW increases by 29.22%, 26.89%, and 50.26% between adjacent temperatures, whereas the average elastic energy storage U exhibits larger increments of 42.67%, 39.26%, and 57.59%, respectively. The disparity in growth rates leads to a cumulative 23% reduction in the loss factor η from the lowest to the highest temperature. This performance variation is governed by the COF, elucidating the influence mechanism of EMPR’s macroscopic mechanical properties from the perspective of material-level friction characteristics. The fundamental cause of the frictional entanglement hysteresis effect (FEHE) during EMPR unloading was identified as the asynchronous reversal of friction directions. A meso–macro linkage between EMPR’s hysteretic behavior and its structural evolution was further established, leading to the development of a Temperature friction hysteresis memory (TFHM) model applicable to microporous topological materials. The model reconstructs full σ-ε curves from a single primary hysteresis loop, enabling rapid generation of high-temperature hysteresis databases for the design of durable, energy-absorbing components.