Distributed edge inference, mobile radio-access-network workloads, and decentralized compute fabrics require stable state handoff under latency, jitter, loss, mobility, and adversarial replay conditions. This paper introduces the Wave-Native Network (WNN), a prototype deterministic phase-coherence overlay that augments conventional network transport with oscillator-state tracking, phase-aware routing, and stability-scored handoff diagnostics. Each node is represented by a deterministic Duffing-type heartbeat state Ψ = ⟨A,θ,ω, ˙x,ts⟩, while routing decisions incorporate latency, jitter, phase mismatch, frequency mismatch, and a scalar stability divergence ∆Φ. The framework is presented as an engineering-level overlay, not as a replacement for TCP, QUIC, TLS, cryptographic authentication, or radio-accessnetwork standards. We evaluate the current WNN prototype using six CSV-based simulation benchmarks: phase-state tracking, jitter and packet-loss robustness, replay-sensitive anomaly scoring, handoff continuity, scalability, and component ablation. Across paired robustness runs, WNN reduces mean phase error by 34.91% ±0.73% relative to the baseline. In the handoff scenario, the baseline continuity score drops from a pre-event mean of 0.945 to a post-event mean of 0.484, whereas WNN maintains a post-event mean of 0.950. Ablation results show that disabling the lock detector increases mean recovery time from 5333.5 ms to 7243.5 ms. Replay detection remains preliminary: the present fixed threshold achieves 76.57% accuracy but only 40.88% recall and 20.12% false-positive rate. These results support WNNas a reproducible phase-coherence testbed for edge synchronization, while emphasizing that deployment-level security and networking claims require further hardware-in-the-loop validation, optimized thresholds, stronger baselines, and broader adversarial testing.