Context
Operator release call (2026-06-11 ~02:10Z): working locked-column drag-and-drop resorting is a hard v13 gate — "before this, i would not dare to release v13." The fix chain starts from a known-good baseline: forensics on #12883 proved src/draggable/container/SortZone.mjs was byte-identical from v12.1 main through dev-pre-#12882 (the multi-body work never touched the motion engine), and the motion/overdrag regressions are pinned to PR #12882's parent-class changes.
The Problem
Post-#12882 dev exhibits: wrong mid-drag header positions ("not even close"), broken overdrag sideways auto-scroll (working a week ago), and is in the suspect set for drop-duplication. Working from this baseline to fix #12883 inverts the difficulty; reverting restores the engine both known-good states shared.
The Fix
git revert b190f60b9 (the #12882 squash) on a feature branch → PR → operator merge. Restores parent SortZone to the v12.1-identical state; the #12880 landing-overshoot returns temporarily and #12880 REOPENS for a re-land that (per the re-land doctrine on #12883) keeps the landing-resolver as the targeting layer, drives the existing visual cadence, and verifies through the #12886 eyes. PR #12881 (grid subclass) is independent and unaffected.
Acceptance Criteria
Out of Scope
The #12883 fix itself (eyes-gated) and #12886.
Decision Record impact
none.
Related
#12883 (conviction evidence), #12880 (reopens), #12886 (gates the re-land verification), PR #12882 (reverted, author notified — concept survives for the re-land).
Release classification: ON the board — first step of the hard release gate chain.
Origin Session ID: 567191c3-16f6-4235-914c-b51fc94d1514
Live latest-open sweep: latest 10 checked 02:10Z, no equivalent. A2A sweep: mailbox zero.
Context
Operator release call (2026-06-11 ~02:10Z): working locked-column drag-and-drop resorting is a hard v13 gate — "before this, i would not dare to release v13." The fix chain starts from a known-good baseline: forensics on #12883 proved
src/draggable/container/SortZone.mjswas byte-identical from v12.1mainthrough dev-pre-#12882 (the multi-body work never touched the motion engine), and the motion/overdrag regressions are pinned to PR #12882's parent-class changes.The Problem
Post-#12882 dev exhibits: wrong mid-drag header positions ("not even close"), broken overdrag sideways auto-scroll (working a week ago), and is in the suspect set for drop-duplication. Working from this baseline to fix #12883 inverts the difficulty; reverting restores the engine both known-good states shared.
The Fix
git revert b190f60b9(the #12882 squash) on a feature branch → PR → operator merge. Restores parent SortZone to the v12.1-identical state; the #12880 landing-overshoot returns temporarily and #12880 REOPENS for a re-land that (per the re-land doctrine on #12883) keeps the landing-resolver as the targeting layer, drives the existing visual cadence, and verifies through the #12886 eyes. PR #12881 (grid subclass) is independent and unaffected.Acceptance Criteria
b190f60b9, CI greenOut of Scope
The #12883 fix itself (eyes-gated) and #12886.
Decision Record impact
none.
Related
#12883 (conviction evidence), #12880 (reopens), #12886 (gates the re-land verification), PR #12882 (reverted, author notified — concept survives for the re-land).
Release classification: ON the board — first step of the hard release gate chain.
Origin Session ID: 567191c3-16f6-4235-914c-b51fc94d1514
Live latest-open sweep: latest 10 checked 02:10Z, no equivalent. A2A sweep: mailbox zero.