This paper compares China's parallel and sequential admissions mechanisms under the dummy-quota policy that limits how many applicants each college can screen. Although neither mechanism is stable or strategy-proof, their weaknesses are structurally different. Under the parallel mechanism, priority violations and profitable deviations are confined to a policy-determined buffer: students routed into colleges' screening pools by exam-score priority but rejected when colleges finalize admissions based on their preferences. The size of this buffer is bounded by the gap between screening and admission capacities. Under the sequential mechanism, by contrast, violations are driven by the round-by-round structure and can expand sharply with congestion and preference concentration. We derive tight upper bounds that characterize these exposures and quantify their magnitudes in calibrated simulations. The results reveal a clear distinction: institutional risk is capped ex ante by design under the parallel mechanism, whereas under the sequential mechanism it scales with market pressure-providing a complementary explanation for China's transition from sequential to parallel admissions.