The INPUT ALIGNMENT delay in ms, 0 when unset — the third head-stage fact, and the one that
had no reader here until #662. Its absence was not a design choice: the lane battery
measured level.alignDelayMs surviving an apply and NOT a cold boot, because the install
replay could not ask for it, so a restored alignment reached the lane on the operator's
NEXT write and VALIDATION had to tell them to re-check alignment after every load.
The canonical head-stage state the install-time push replays (#122) — read from the software mixer, whose per-strip records the session restore has already written by install time.
FADER and PAN are absent by construction, for the reason spelled out on installHeadStageNativeMirror: their C fields have exactly one writer, and it is not this bridge. A seam that could not name
setFadercannot grow a second one back.