Row-repaint closeout against the persistent combined source

Audit date: 2026-10-01. This is a source/evidence reconciliation; no board, serial port, build slot, or runtime was used by this audit.

Existing matched pair

../row-repaint/pair/README.md records physical injected-input runs for both picker rows, six swipes per row in each arm, with separate restoration and identities. It compares the previous Rust client from source b57ba41ccf6755c54039fc85b0a54145001b186a against candidate source 42e22d4353e35d9c4f6367ed5b159c80b489ec5d and candidate binary /nix/store/2mprr7lqh4hznwz9z1zmv7vamf5nbnna-k230-shell-rust-riscv64-unknown-linux-gnu-0.1.0. For that pair, active presentation-gap median improved from 134.121 ms to 76.641 ms and p95 from 210.765 ms to 153.287 ms. It also records that the selected preview differed between the two arms, cadence/jank was not resolved, and the candidate was restored afterward. This is injected-touch performance evidence, not real-finger acceptance.

Current installed source

The committed docs/evidence/boot-verification/coherent-ordinary-boot/README.md and postboot.json identify the persistent system as /nix/store/p1a1hz9n8s4g8qyr55ffl3dzbnjgqwr8-nixos-system-nixos-26.11.20260919.20b1ddd, the running Rust executable as /nix/store/3hy6h165ii649z6vjzjd36jwg16d37rc-k230-shell-rust-riscv64-unknown-linux-gnu-0.1.0/bin/k230-shell-rust, and the installed application source as 5bb67db128210f830dab4de0d20a4b0eca13c578. The ordinary reboot record shows the selected system and profile agree, the shell services are active, and the saved appearance generation/report hash matches before and after boot. The record explicitly leaves real-finger verification false.

The current source is a descendant of the measured row-repaint change, but the pair does not identify this exact combined Rust binary. The read-only comparison command was:

git diff --stat \
  42e22d4353e35d9c4f6367ed5b159c80b489ec5d..5bb67db128210f830dab4de0d20a4b0eca13c578 \
  -- nix/rust-shell-client/src/theme_ui.rs \
     nix/rust-shell-client/src/theme_carousel.rs \
     nix/rust-shell-client/src/main.rs \
     nix/rust-shell-client/src/render.rs \
     nix/rust-shell-client/src/background_decode.rs \
     nix/rust-shell-client/src/theme_thumbnails.rs

main.rs and render.rs changed between the measured and installed source identities; render.rs includes changes in variant_size and paint_carousel. The pair therefore cannot establish the current build’s render or presentation cost. The ordinary reboot proves persistence and saved appearance only. No new task is marked complete by this audit. Task 17.3 still needs a matched baseline/current-candidate injected comparison on the reserved board, with independent restoration and the existing identity/capture rules. Task 14.4 real-finger acceptance, task 10.7 picker fit and tap timing, and the wider 13/15 gates also remain open.