Say that OTA releases are asked for, because they are master
Three places promised that landing fedac/native/ on main starts an oven build. None of them do. The native poller reads NATIVE_POLL_INTERVAL_MS with a default of 0, nothing in the repo sets it, and the disabled branch logs that the manual trigger is the way in. Nothing is broken here: papers polls every 60s while native and recap sit at 0, so two of three being off is a decision, and auto-shipping every native commit to real hardware is a strong thing to do by accident. What the promise cost was invisible. Push, wait, and the published OTA is simply whatever it was — no error, no queue entry, just a stale kernel that flashes onto a stick and boots fine. Today that was a five-week-old build about to go onto a USB to test a change made this morning. The poller's own header claimed a 60s default while the line under it read 0, which reads like it was copied from the papers one and never revisited. It now says why it is off and what turning it on would mean.