Split tier2 kind-8 setup bridge fields

artifacts/exports/rt3-1.06/runtime-effect-kind8-tier2-setup-core-note.md

@@ -85,6 +85,9 @@ One narrower outlier still survives that wider scan:
 - no other bundled `rt3_105` map matches the full current `Louisiana.gmp` setup-core tuple
   `payload_word_0x14 = 1870`, `payload_byte_0x20 = 0x3a`, `payload_word_0x3b2 = 2`,
   `candidate_header = 0xcdcdcdcd / 0xcdcdcdcd`
+- the two fields on the currently grounded Tier 2 bridge split differently in the wider corpus:
+  - `payload_byte_0x20 = 0x3a` is unique to `Louisiana.gmp`
+  - `payload_word_0x14 = 1870` has only one peer, `Mexico.gmp`
 - even the nearest header-class peers still diverge:
   - `Argentina Opens Up.gmp`: same header pair, but `payload_word_0x14 = 1880`,
     `payload_byte_0x20 = 0x57`, `payload_word_0x3b2 = 1`
@@ -142,6 +145,28 @@ That means the next upstream question is now smaller than before:
 - or whether the remaining differentiator still sits elsewhere, most likely in the sparse
   recipe/runtime side rather than the setup-panel-only `+0x3b2/+0x3ba` pair
 
+## Adjacent setup-launch cross-check
+
+One adjacent setup-side comparison also stays divergent rather than collapsing the `Louisiana`
+outlier into one easy peer group:
+
+- `compare-setup-launch-payload`
+  over `Louisiana.gmp`, `Mexico.gmp`, `Argentina Opens Up.gmp`, `Crossing the Alps.gmp`,
+  `Spanish Mainline.gmp`, and `Chicago to New York.gmp`
+  shows no shared launch-flag or token-block signature with `Louisiana.gmp`
+- `Louisiana.gmp` carries
+  - `launch_flag_byte_0x22 = 0x8f`
+  - `launch_token_block_0x23_0x32 = 0012918f000000000000000000000000`
+  - nonzero selector values only for `Central Pacific = 145`, `Germantown = 18`,
+    `Texas Tea = 143`
+- all five comparison maps diverge on that launch-side signature, even when they are near peers on
+  the setup-core or candidate-header side
+
+That does **not** yet make the launch-token band part of the grounded Tier 2 bridge. Current atlas
+evidence still only grounds the direct Tier 2 bridge through setup payload `+0x14/+0x20` into
+`[profile+0x77/+0xc5]`, then through `0x00436d10 / 0x00443a50`. But it does make the upstream
+picture less likely to collapse to one trivial peer class before that bridge.
+
 ## Current implication
 
 The upstream side of the Tier 2 strip is now bounded more tightly:

docs/rehost-queue.md

@@ -577,7 +577,17 @@ Working rule:
     `[profile+0x77/+0xc5]`, while `+0x3b2/+0x3ba` still stop on the setup-panel threshold or
     scroll path. The next pass should therefore test whether the unique `Louisiana.gmp`
     `+0x14/+0x20` pair is enough to explain its Tier 2 runtime shape through that grounded bridge,
-    or whether the sparse recipe/runtime side is still the more plausible differentiator.
+    or whether the sparse recipe/runtime side is still the more plausible differentiator. One
+    adjacent `compare-setup-launch-payload` check now sharpens that too: `Louisiana.gmp` does not
+    collapse into the nearest setup-core peers on the launch-token side either, but that band is
+    still not part of the currently grounded Tier 2 bridge. So the next pass should keep launch
+    tokens as supporting context while testing the already-grounded `+0x14/+0x20 -> [profile+0x77/+0xc5]`
+    bridge against the sparse recipe/runtime family. The same widened setup-core note now shows the
+    split inside that bridge too: `payload_byte_0x20 = 0x3a` is unique to `Louisiana.gmp` across
+    all 41 bundled `rt3_105` maps, while `payload_word_0x14 = 1870` has only one peer (`Mexico`).
+    That makes the campaign/setup-byte side of the bridge the strongest remaining setup-core
+    differentiator, with the year-word side a secondary companion rather than a full peer-group
+    collapse.
     kinds”; it is the smaller set of scenario-specific records where that sweep explicitly writes
     `[event+0x7ef]` itself or a still-later owner does.
   - two explicit trigger-kind materializations are now grounded inside that retagger: