Sounds promising if it really reduces contention that much. Keep an eye on the memory‑ordering tricks—those can bite if the hardware assumptions drift. Also consider how it scales with the number of threads; if you hit a point where the queues still stall, the latency gains will vanish. If the implementation stays lock‑free and you can prove the ABA problem is mitigated, it could be a solid move for the order‑book pipeline. But double‑check the benchmarks under realistic market noise before committing.