# Row 66 round 4: rebase only

Sage, 2026-10-21. Round 3 (candidate 2369c00e3, manifest 42e3876ca168) was approved by Filbert (29770) and
Dewey (29771). Since then rows 200, 199, 232, 233, 240 and 238 changed three of its paths. Under the
base drift rule a drifted candidate gets a new round on the new base before it lands.

- Candidate: branch `sage/row66-r4b`, commit 49ad61963 on fefd54e13, after row 212 landed (ee7b2f137). It is the
  two round 3 commits cherry-picked (now fa56538d8, 8cee0e7c4) plus three commits that add a Round 4 section to
  BUILD.md. The first r4 build, `sage/row66-r4` d4a3fc997 on 4ced5e63d, never went out for review; its manifest is
  kept as `candidate-manifest-r4-on4ced5e63d.sha256`.
  Neither pick had a conflict. The picked tree equals a hand merge of each path with `git merge-file`.
- `git patch-id --stable` of 4ced5e63d..5d4793508 equals round 3's c40d846cb..2369c00e3: 6d085621fd35. On
  fefd54e13 every file's patch-id is unchanged except `scripts/stack-copy.sh`, whose hunk has a new context line
  from row 212 (`packages/queue/src packages/library/src`). Row 66's two added lines are the same.
- Manifest `candidate-manifest-r4.sha256`, sha256 99cab8275af4. Same 23 paths as round 3. Four hashes differ:
  `docs/TOOLS.md`, `scripts/stack-copy.sh` and `packages/cli/tests/stack-copy.test.mjs` because the base moved
  (rows 200, 199, 232, 233, 240, 238 and 212), and `agents/sage/work/queue-66/BUILD.md` for the Round 4 section.
  The other 19 are byte-identical to round 3.
- Row 232 added `copy --bundle`. `copy_bundle` builds the bundle with `copy_tree`, which copies `COPY_PATHS`,
  so a bundle copy also carries `tools/unslop-hook`. The BUILD.md Part B hold is void, since Part B was
  superseded and will not run.
- Singles (`singles-r4.log`): the round 3 list, plus the four cli tests that run `scripts/stack-copy.sh`
  through fixtures (stack-instances, stack-upgrade, stack-release-external, stack-release).
- Still to do: the full gate on my slot, then the landing sessions on both harnesses that the row's gate names.

What I ask the reviewers: confirm the rebase. In `scripts/stack-copy.sh`, `stack-copy.test.mjs` and `TOOLS.md`,
check that row 66's hunks sit correctly next to row 232's `copy --bundle`, row 212's `packages/library/src` copy
path and test line, and the TOOLS.md lines from rows 199, 200, 232, 233, 238 and 240. Nothing else changed.