Inventory management of seasonal products with uncertain quality involves challenging decisions under simultaneous demand and quality uncertainty, while most existing approaches, after inspection, still rely on a rigid binary “accept-or-return” rule. This study develops a three-state threshold newsvendor model for a seasonal perishable product, in which inspection errors, customer returns, logistical limits on replacement, and two-channel discount-sensitive demand are incorporated in an integrated way, and the supplier’s participation constraint is explicitly enforced in designing the optimal policy. Owing to nonlinearity and the presence of min/max operators, the expected profit function is not analytically tractable; therefore, a combined simulation–optimization framework is adopted, where expected profit is estimated via Monte Carlo simulation and the optimal policy is obtained using Bayesian optimization and benchmarked against standard metaheuristics. Numerical results show that, relative to the accept/return policy, activating the replacement mechanism within the three-state framework increases the retailer’s expected profit by about 14.3% and the supplier’s profit by about 6.2%, while simultaneously reducing shortages and waste significantly. Scenario analysis under critical conditions indicates that model performance is robust to parameter changes and that managed replacement acts as an effective risk-hedging instrument. Sensitivity analysis further reveals that, in addition to purchase and selling prices, operational variables such as replacement effectiveness and inspection accuracy have a direct impact on shortages and waste. These findings suggest that moving from traditional full-return contracts toward “smart” contracts embedding a capped replacement mechanism can substantially improve both profitability and operational performance in seasonal perishable supply chains.