docs: close pair adaptation preview

CURRENT_CHECKPOINT.md

@@ -28,7 +28,9 @@ P9.0R9A baseline: `1ce563b401dc9d6091ea54044cac6a838c27c283`
 P9.0R9A product preview source: `32b4fdb`
 P9.0R9A Gallery item: `VL8pU8RUhXNn`
 P9.0R10 baseline: `d8b359a12f5641257c549894a8114245fa987962`
-Stage: `P9_0R10_PAIR_ADAPTATION_ACTIVE`
+P9.0R10 product preview source: `41204acafa911a505f5b24a493aa032983e706d1`
+P9.0R10 Gallery item: `gHS8jEZdGa2x`
+Stage: `P9_0R10_PAIR_ADAPTATION_H5_PREVIEW_READY`
 
 P9.0R10 owns the next compatibility boundary after automatic formation. All six
 legal two-Recruit pairs must receive a truthful first-Victory diagnosis, choose
@@ -37,7 +39,23 @@ member behavior in battle two, and reach a causal second result. The exceptional
 Hearthguard/Wayrunner path must retain the Runner flank counter instead of
 falling through to healing. R10 stops at the second-result page; later R5-R7
 systems, save/content/release scope and authoritative combat remain frozen.
-Shortest next action: implement and verify the six-pair by two-equipment matrix.
+Clean/upstream source `41204ac` now carries all six pairs through truthful
+first-Victory diagnosis, the unselected safety-versus-tempo loan and the second
+result. The 12-path matrix passes exact HP `90/100` versus `70/82`, times
+`7.05/6.00`, Coin 20, ranged/healing/flank-counter identity and exceptional-
+pair Home cleanup. Parsed touch reports `touch=5 fallback=0 direct=0`; the
+serial R1-R10 run passes 22 tests with strict markers zero. Eight `844x390`
+normal-render states pass 44px bounds after repairing an English-130% overflow.
+
+The fixed-source runtime-only Web build is
+`runtime/build/web-preview/41204ac-r10-fast/`; `export.sha256` digest is
+`7aef140b95d4c5d5bcc5c5815f4ae02894ee49367a6fbeb87dc07915d258f859`.
+PCK inventory is `227` files with forbidden paths zero; Gallery item
+`gHS8jEZdGa2x` serves a byte-identical PCK and reaches the corrected Chinese
+Title with browser diagnostics zero. Remote cold start remains about four
+minutes. Shortest next action: Owner plays the new H5 and judges whether every
+pair's result and safety-versus-tempo choice are understandable and worth
+continuing.
 
 Owner review rejected the R9 interaction boundary: players should select two
 people, not manually perform separate front and rear slot selection. R9A owns