Light catch-up is opaque: no warning that releasing manual hold turns the lamp ON, no way to cancel a catch-up, and no stated next scheduled light change

OPERATOR-RAISED 2026-09-10. Three gaps around light catch-up. 1. NO WARNING ON HOLD RELEASE. Read from the firmware: manual_light_hold self-heals only when photo_was_want_on && !want_on. During catch-up want_on is TRUE, so clearing the hold does NOT leave the lamp where it is — it immediately drives the SF1000 to target brightness and keeps it there until the debt is repaid or the min-dark floor is hit. The SPA says nothing about this; the hold card only explains that the hold freezes the lamp. An operator releasing a hold at night could restart the lamp without expecting it. 2. NO WAY TO CANCEL A CATCH-UP. There is no operator-facing control to forgive the debt. The mechanism exists internally — photo_delivered_s = target (“freeze paid”, used when the 2×4 goes inactive at line ~5066) — but nothing exposes it. Turning off auto_photoperiod_enabled stops catch-up but kills the whole schedule AND leaves the debt, so re-enabling resumes it. A real cancel therefore needs a FIRMWARE control and a hub flash; it cannot be done from the brain or SPA alone. 3. NEXT SCHEDULED LIGHT CHANGE NOT STATED. While catch-up runs, the desk shows CATCH-UP ACTIVE but not when it is expected to end (debt remaining vs min-dark floor) nor when the next lights-on is due. The clock chip shows OFF IN / ON IN for the NOMINAL window, which is misleading during catch-up because the lamp will run past nominal off. SPLIT: (1) and (3) are pure SPA and ship without a flash. (2) needs firmware, so the control must render as an INVENTED-style affordance (dashed, ‘possible after the next hub flash’) until the entity exists — never a dead button.