Rig-gated - the studio-facing surface

Two things no gate in this repository can see: a real controller and a real show. Extracted from Plan 0158 at that plan’s close (2026-09-10) so the plan could close on its six completed dev phases without waiting on a projector session, exactly as every item above was.

Nothing in CI sends a datagram from another program - the suite’s loopback test is the player talking to itself - and nothing in CI watches a projector for ten minutes. The frame-time figures Plan 0158 Phase 6 recorded are a debug build on one machine over ~7 s, which is a shape rather than a reading; the number this item exists to produce is the release one, at the projector.

• Drive the player from a real OSC sender. From TouchDesigner, or any sender on the same machine, run ritmolux --control 127.0.0.1:9000 --events against a playing preset. Confirm ctl/param moves a parameter and ctl/preset switches the preset, and that health’s counters acknowledge it - ctl_rejected staying flat is the half that says the sender and the decoder agree on the wire, not merely that something happened. The vocabulary is in the studio control protocol spec; the flags are in Configuration. (Plan 0158 Phase 7, extracted at that plan’s close 2026-09-10.)
• Read the preview’s real cost on the projector. A windowed player on the projector with --preview stdout piped to a file, playing a track for ten minutes. Confirm the preview frames are visibly the show, then record the release-build frame_ms p50 and p99 from the preview line with the readback on and off here. The debug-build pair to beat is p50 37.18 / p99 76.44 on and p50 30.72 / p99 76.24 off - ~6.5 ms at the median and nothing at the p99, on a build whose show already ran at 30 fps. A run that delivers zero frames is not a reading: the frame count on the exit line is ADR-0172’s witness that the readback actually ran. (Plan 0158 Phase 7, extracted at that plan’s close 2026-09-10.)