Shock-tube experiments are conducted to investigate the Atwood-number dependence of hydrodynamic instability induced by a strong shock with a Mach number exceeding 3.0. The compressible linear theory performs reliably under varying compressibility conditions. In contrast, the impulsive model significantly loses predictive accuracy at high shock intensities and Atwood numbers ($A_t$), particularly when specific heat ratio differences across the interface are pronounced. To address this limitation, we propose a modified impulsive model that offers favourable predictions over a wide range of compressibility conditions while retaining practical simplicity. In the nonlinear regime, increasing $A_t$ enhances both the shock-proximity and secondary-compression effects, which suppress bubble growth at early and late stages, respectively. Meanwhile, spike growth is promoted by the spike-acceleration and shock-proximity mechanisms. Several models reproduce spike growth across a wide range of $A_t$, whether physical or incidental. In contrast, no models reliably describe bubble evolution under all $A_t$ conditions, primarily due to neglecting compressibility effects that persist into the nonlinear regime. Building on these insights, we develop an empirical model that effectively captures bubble evolution over a wide $A_t$ range. Modal evolution is further shown to be strongly affected by compressibility-induced variations in interface morphology. The effect is particularly pronounced at moderate to high $A_t$, where it suppresses the fundamental mode growth while promoting higher-order harmonic generation.