2026-10-25: camera idle zoom regression checked on current QA build

Current private EXE SHA256:
ff4fdaee69ba2fda45d809eca4e52bf7b3c9718ac0a93cf38668f1b31d797ff4

The normal original-menu Battlefield input replay used Yoshi, moved with the
left stick, and stopped. In the 17 sampled camera traces from ticks3500..3992,
player state is idle St_Wait_Main and camera-to-target distance remained
794.98 world units (min=max). The current regression does not reproduce the
reported zoom-in while stopped. Yoshi ends alive/in level6 with normal control;
isolated slot/minigames saves match their seed. Process exits0. Audio output was
disabled for this test. Screenshots3100/3500/3900 show gameplay during movement
and after stopping. Evidence is private at
work/native-movement-AnalogFull-off-20261020-resume-camera-idle-20261025;
camera-idle-acceptance.json records the measured result.

This selected Battlefield case does not test doors, cap pickup, physical Switch
Pro input, every camera zoom preset, or the user's installed F: executable.
The F: installation was not changed. The existing lens correction was only
regression-tested; no code changed in this camera test.

Next difficult target from the user's open reports: ordinary castle-to-Secret-
Slide entry and missing finish/timer UI. Earlier two route pilots fell from the
narrow slide and failed to reach the native Star box; no gameplay fault was
proven. The evidence-led route progress/pilot repair needs Astra Low, as the
existing handoff recommends. After that investigate the slide's genuine finish
and timer using original objects and rules. Keep runtime audio muted and show
screenshots. The Koopa pickup/save/reload/repeat lifecycle has since been
verified separately; older Koopa-open notes are historical.
