This study examines algorithmic peak-hour delivery restrictions as a form of digital platform governance and evaluates their welfare implications. Motivated by recent practices in South Korea's highly concentrated food delivery platform market, we develop a stylized model in which a dominant digital platform chooses an endogenous delivery service radius to restrict delivery service areas during peak demand periods. The platform trades off congestion relief and operational performance against reductions in consumer choice and small business access. While algorithmic service area restrictions can improve delivery times and reduce order cancellations by alleviating congestion, they simultaneously shrink consumers' effective choice sets and contract restaurants' accessible markets. Because the platform internalizes the operational benefits of congestion relief but not the welfare losses associated with reduced choice and access, the privately optimal service radius is smaller than the welfare-maximizing radius in the model. Our findings highlight that platform control operates not only through prices but also through rule-based mechanisms that govern access. As a result, seemingly technical operational policies can generate welfare distortions by altering the feasible set of market interactions. We also suggest that performance-based regulation can improve welfare when it conditions the permissibility of access restrictions on demonstrated performance necessity.