Real historical replay: timeline pins load run's own checkpoint GLBs

- renderTimeline rebuilds #timeline-track after each snapshot and wires
  click-to-pin on completed checkpoints (kept placeholder nodes otherwise)
- selection loads the pinned run's own steps/<n>-*.glb (generation/
  ownership guards apply), pins survive live snapshots, 跟随最新 restores
  live scope, replay pauses/scrubs through 已完成 versions
- placeholder fallback now clearly labeled (占位…) only before any run
- acceptance: historical_replay_regression (9 checks) pins mid-run,
  asserts no auto-follow strays, late-older-response precedence, served
  checkpoint sha256 vs CLI reference, and viewport geometry change
- fix: timeline-track id missing in index.html crashed renderTimeline
  (null firstChild), leaving static placeholder nodes non-clickable
- driver: 71/71 checks, zero console/page errors

1 files changed, 18 insertions, 10 deletions

diff --git a/README.md b/README.md
index 470dc8b..977c984 100644
--- a/README.md
+++ b/README.md
@@ -188,7 +188,7 @@ dotnet run --project src/SomhairlesDream.Cli -- run \
   reference run; a failure run (pre-existing run directory) shows status
   `失败` with the error surfaced in `run-message`; zero console errors on
   all pages.
-- Browser acceptance driver (see above), 61/61 checks green on desktop and
+- Browser acceptance driver (see above), 71/71 checks green on desktop and
   narrow viewports: framing asserted numerically from the projected scene
   bounding box (not by trusting whole-canvas hashes alone), frozen
   screenshots byte-deterministic, GLB hash equality, failure surfacing,
@@ -196,16 +196,24 @@ dotnet run --project src/SomhairlesDream.Cli -- run \
   responses and fulfills them late, reordered, or with errors across nine
   runs, asserting via actual scene geometry (`THREE.Object3D` types/colors)
   and the visible note that superseded runs' responses never repaint the
-  viewport, same-path reruns still reload, and run switches to empty state
-  fall back to the replay frame. It also proves: a failed current
-  latest-step load (after run completion) shows a visible retry affordance
-  on the note and a user click re-requests the same owner/path without
-  SSE reconnect; within one run an older step resolving after the newer
-  step is ignored; and a live step response still pending after entering
-  replay never repaints the selected replay frame view. Held responses
-  are drained and the route handler removed before the page closes. The
+   viewport, same-path reruns still reload, and run switches to empty state
+   fall back to the replay frame. It also proves: a failed current
+   latest-step load (after run completion) shows a visible retry affordance
+   on the note and a user click re-requests the same owner/path without
+   SSE reconnect; within one run an older step resolving after the newer
+   step is ignored; and a live step response still pending after entering
+   replay never repaints the selected replay frame view. A historical
+   replay regression pins a real checkpoint mid-run by clicking its
+   timeline node (the request carries the run's own projectId/runId/path),
+   keeps the pin through later live snapshots (no auto-follow reloads),
+   returns to live scope via 跟随最新 with the newest checkpoint, ignores a
+   late older historical response after a newer selection, and asserts
+   both the served checkpoint identity (sha256 equals the CLI reference
+   GLB) and that selecting a different version changes the actual viewport
+   geometry (THREE.Box3 rel diff ≥ 0.05). Held responses
+   are drained and the route handler removed before the page closes. The
   frontend refits the camera to the loaded model
   (`THREE.Box3`-based framing) whenever the canvas resizes, and a resize
   never reverts the viewport to replay placeholder geometry.
 - NOT verified: production deployment of the old site remains untouched;
-  `--render` PNG renders; replay button flow against a historical run.
+  `--render` PNG renders.