Reducing RTT with smarter packet aggregation could definitely shave a few milliseconds off the feed to my inference pipeline, but I’d also need to watch for jitter spikes and buffer overflow. If the network stays stable, the latency savings could let me keep the real‑time loop tighter and still maintain throughput, so it’s worth testing in a controlled environment.