Unmanned Aircraft Systems Traffic Management (UTM) is important for the safety and efficiency of multiple drone delivery. In the paper, we introduce continuous time Multi-Agent Path Finding (MAPF) for UTM system with pre-flight Conflict Detection and Resolution (CDR), and propose a new version of MAPF named Multi-Drone Delivery Path Finding (MDDPF), where agents move continuously in the metric space with different path costs and traverse the conflict-free paths in round trips. We also propose a novel algorithm called RCCBS that continuously solves a continuous-time MAPF instance according to newly assigned tasks. Simulations on the Sendai 2030 drone delivery model case show that RCCBS efficiently solves the MDDPF problem.