IP-Safe Routing Benchmarks for a Protected Modular Qubit Interconnect: Nonlocality Stress Tests, Compiler Cross-Checks, and Baseline-Specific Trade-offs

Quantum processors pay a routing cost when a logical circuit requests two-qubit interactions that are not adjacent on the hardware connectivity graph. This preprint reports an IP-safe, fixed-seed, topology-aware transpilation benchmark of FPN Hub, a proprietary deterministic modular qubit-connectivity architecture whose construction rules, edge lists, coupling maps, generator logic, hub-selection rules, and module internals are not disclosed. The central claim is deliberately narrow: FPN Hub is not universally superior, but one anonymized sparse inter-module configuration (Variant A) shows a reproducible routing advantage for fanout and nonlocal-interaction workloads, whereas a second anonymized configuration (Variant B) degrades as nonlocality increases. The strongest evidence is a controlled nonlocality stress test in which an otherwise fixed circuit is swept from local to fully long-range interactions. Under Qiskit, the Variant B/Variant A cost ratio changes from slight Variant B advantage at f = 0 to strong Variant A advantage at f = 1; a directional tket cross-check reproduces the high-nonlocality ordering, although the exact emergence curve is router-dependent. Additional MQT Bench checks show workload selectivity rather than universal dominance. The study is a transpilation-level routing benchmark, not a hardware validation or FTQC/QEC proof.