fix(showcase): fence-kit hero shares one post at each joint

The three-section hero placed whole copies of FenceLow at the 1.60 m tile
pitch, so every joint showed two posts band to band. The leading section
is still the checked mesh. Each follow-on section is now a render-only copy
with its first post group (post, cap, shoe, bands, found as mesh shells)
removed, stepped at the post pitch 2*HX, so its rails and brace tenon into
the previous section's trailing post. The asset, its budgets and its
falsifiers are unchanged; the asset sheet still judges FenceLow alone.

Evidence (live, local):
- .scratch/blender-5.2.1-windows-x64/blender.exe (Blender 5.2.1 LTS):
  default exit 0, base_tris=1012 r1=0.5000 r2=0.2174 collider 24; --output
  framing fill x=0.837 y=0.667, margins >= 0.081.
- .scratch/blender-4.5.11-windows-x64/blender.exe (Blender 4.5.11 LTS):
  default exit 0, base_tris=1012 r2=0.2194 export 75900 B.
- tests/check_falsifier_targets.py --run on both binaries: every fence-kit
  falsifier returns its documented code (9, 15, 16, 16, 17, 18, 19, 20, 21).
- Contact sheet: luma 0.227 / stage 0.214 / warmth +0.134 / sat 0.180.
- Asset sheet: FenceLow unchanged, in the lineup.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Author: TMHSDigital

docs/gallery/asset-sheets/fence-kit.webp

@@ -17,13 +17,18 @@ rail and housed 12 mm into both posts. It clears the bottom rail at the left
 post and the top rail at the right one. Rails and kick tenon into the post
 volume.
 
-**Hero.** The still shows the kit doing its job: three linked duplicates of
-the one checked mesh at the 1.60 m pitch, so every joint in the picture is
-the tile fit the budgets assert — the paired posts meet band to band,
-`KIT_CLEAR` apart. The run stands in a render-only packed-earth strip, and
-the wood weathers in broad silver-grey patches (object-space noise) over
-the per-member tone and grain. The strip and the duplicates are staging;
-the asset sheet judges the single `FenceLow` section.
+**Hero.** The still shows a three-section run that reads as one fence,
+with one post at each joint. The leading section is the checked `FenceLow`
+mesh. Each follow-on section is a render-only copy with its first post group
+(post, cap, shoe and bands, found as mesh shells) removed, stepped at the
+post pitch `2·HX`. Its rails and brace then tenon into the previous
+section's trailing post. Placed whole at the 1.60 m tile pitch, the sections
+meet band to band, which is the tile fit the budgets assert, but every joint
+shows a doubled post. The earlier hero did exactly that. The run stands in a
+render-only packed-earth strip, and the wood weathers in broad silver-grey
+patches (object-space noise) over the per-member tone and grain. The strip
+and the trimmed copies are staging; the asset sheet judges the single
+`FenceLow` section.
 
 It asserts **budget conformance** of the generated result. It does not
 witness an API contract. "It rendered without error" is not a check. This

docs/gallery/assets/fence-kit-hero.webp

@@ -1262,19 +1262,53 @@ def render_still(low, wood, tex, path, engine):
             ob.hide_render = True
             ob.hide_viewport = True
 
-    # The hero is the kit doing its job: three sections at the tile pitch,
-    # linked duplicates of the one checked mesh, so every joint in the
-    # picture is the tile fit the check asserts (posts meet band to band,
-    # KIT_CLEAR apart). A run parent carries the yaw; level on the floor,
-    # since an X tilt sinks one face of the shoes.
+    # The hero is a run that reads as one fence, not three panels parked end
+    # to end: the leading section is the checked mesh itself, and each
+    # follow-on section is a render-only copy of it with its first post
+    # group (post, cap, shoe and bands, found as mesh shells) removed,
+    # stepped at the post pitch 2*HX so its rails and brace tenon into the
+    # previous section's trailing post - one post per joint. Placed at the
+    # 1.60 m tile pitch instead, whole sections meet band to band (the tile
+    # fit the check asserts) but every joint shows a doubled post. A run
+    # parent carries the yaw; level on the floor, since an X tilt sinks one
+    # face of the shoes.
     run = bpy.data.objects.new("FenceRun", None)
     scene.collection.objects.link(run)
     run.rotation_euler.z = math.radians(-24.0)
+    follow_me = low.data.copy()
+    follow_me.name = f"{low.name}.Follow"
+    bm = bmesh.new()
+    try:
+        bm.from_mesh(follow_me)
+        post_edge = POST_XS[0] + POST_W / 2.0 + IRON_T + 0.002
+        seen = set()
+        doomed = []
+        for v in bm.verts:
+            if v in seen:
+                continue
+            group, stack = [], [v]
+            seen.add(v)
+            while stack:
+                cur = stack.pop()
+                group.append(cur)
+                for e in cur.link_edges:
+                    o = e.other_vert(cur)
+                    if o not in seen:
+                        seen.add(o)
+                        stack.append(o)
+            if max(g.co.x for g in group) < post_edge:
+                doomed.extend(group)
+        bmesh.ops.delete(bm, geom=doomed, context="VERTS")
+        bm.to_mesh(follow_me)
+    finally:
+        bm.free()
+    pitch = 2.0 * POST_XS[1]
     sections = [low]
-    for i in (-1, 1):
-        dup = bpy.data.objects.new(f"{low.name}.Run{i:+d}", low.data)
+    low.location.x = -pitch
+    for i in (0, 1):
+        dup = bpy.data.objects.new(f"{low.name}.Run{i}", follow_me)
         scene.collection.objects.link(dup)
-        dup.location.x = i * TILE
+        dup.location.x = i * pitch
         sections.append(dup)
     for ob in sections:
         ob.parent = run