nx_video_qoe_live.nx

nx_video_qoe_live.nx -- LIVE video-QoE measurement (operator 2026-07-03: "provide fps and stability and other measurements ... too many false positives that our video chat is really good ... honest feedback if we need more hardware"). NOT a simulation: two real TLS-1.3+WSS sessions through the PRODUCTION edge (nishifamily.com -> sites.elf /signal pump -> nx_signaling_v2 relay), sender A paces REAL room-framed video-kind frames (vc_wire_pack kind 0x56, 6000B payload = a realistic ladder JPEG) at a target cadence, receiver B timestamps arrivals. Ping-pong (one frame in flight) = backpressure-safe on the production relay. Measures PER RUNG of a load ramp (20/40/80/160 fps): achieved fps, path latency p50/p95/max, late>150ms (playout-deadline class, the intl_h2h convention), gap>500ms (hard-freeze class), and a connect/TLS setup time. The ramp knee = the honest hardware ceiling of the CURRENT edge. HONESTY BOUNDS (printed, so no false positives): this measures the Texas-side LAN->edge->LAN path ONLY; the Minsk leg needs a probe RUN FROM MINSK (registered follow-up) or calibration vs RIPE-Atlas real-path data (banked by nx_vqoe_research_fetch). Log -> knowledge/status/video_qoe_live.log expect_exit: 0 (measurement SOUND -- controls hold; the numbers themselves print honestly) license_tier: ORIGINAL