Abstract: Sharp leading edges are widely used in near-space and ultra-low-Earth-orbit hypersonic flight, but unavoidable surface imperfections can produce small windward cavities and gaps that locally distort the near-wall flow and alter the resulting aerodynamic loads. This work employs the Direct Simulation Monte Carlo (DSMC) method to examine how a windward micro-cavity on the sharp leading edge of a two-dimensional wedge alters aerodynamic heating and force loads. Simulations are performed at three altitudes (70, 110, and 150 km) with the cavity depth varied from 0.05 to 2 mm. The cavity-resolved temperature, velocity, and pressure fields are obtained. Distributions of the surface heat-flux, pressure, and shear-stress coefficients along the cavity-perimeter coordinate are evaluated, and incident velocity statistics of gas particles on the cavity bottom are used to characterize changes in the incidence state. At 70 km, the cavity has little effect on the bow-shock shape and stand-off distance, but it promotes a closed recirculation region with low-temperature gas retention, markedly reducing bottom heating; accordingly, the bottom heat-flux coefficient decreases monotonically with increasing cavity depth. At 110 and 150 km, stronger rarefaction lowers the overall temperature and pressure levels and weakens the sensitivity of circumferential heating to cavity depth. Meanwhile, the internal recirculation becomes more distinct and a low-pressure band develops along the sidewall, and deeper cavities induce more pronounced disturbances in the near-field velocity around the leading edge. Particle statistics further show that, in the highly rarefied regime, the streamwise incident-velocity distribution on the cavity bottom changes from a single-peak form to a bimodal form consisting of a freestream high-speed peak and a low-speed peak, with the low-speed contribution increasing as the cavity deepens. These results support aerodynamic assessment and design of sharp-leading-edge micro-features under rarefaction effects and provide incident conditions derived from particle statistics for gas–surface interaction modeling.