@@ -2672,3 +2672,23 @@ Seven review rounds on the forward guard that skips files patching `Deno.*` (for
 - Commit "tools: codemod-deno-ral --reverse reports canonicalPathSync instead of mapping it to Deno.realPathSync": `--reverse` no longer maps `canonicalPathSync` to `Deno.realPathSync` (the monkey-patched `normalizePath` on stable branches); it reports the binding in `left` with a hint naming `originalRealPathSync`, before the re-export check so a re-exported name keeps the hint. Pinned by the `reverse-realpath` and `reverse-realpath-reexport` fixtures. TDD: the new test failed first with a type error (`kCanonicalReverseHint` not exported). Suite after: `ok | 28 passed | 0 failed`. CLI dry-run on the fixture printed `:2: left: 'canonicalPathSync' (deno_ral/fs.ts) is not Deno.realPathSync on branches that load core/deno/monkey-patch.ts; use originalRealPathSync from deno_ral/original-real-path.ts; edit by hand` and `Summary: 1 files scanned, 1 would be rewritten, 1 replacements, 0 with collisions, 0 skipped, 1 left`. Plan text updated at Behaviour (reverse), the Task 5 backport text and the review checklist.
 - Commit "tools: codemod-deno-ral freezes the write guard and skips the known patch files by path": the write guard is documented as best-effort and frozen, and `kCodemodPatchFiles` (`src/core/deno/monkey-patch.ts`, `src/core/performance/metrics.ts`) is a backstop checked after the guard, so a caught file keeps the guard's reason. Pinned by `patch-file-backstop.input.ts` (a `Reflect.get` write the guard misses) run under both listed paths. TDD: the test failed first with `result.skipped` `undefined` (the file was rewritten). Probe before documenting the gaps (Deno 2.7.14, `codemodFile` dry run): all three reproduced — `Reflect.get(globalThis, "Deno")` → skipped undefined, 1 replacement; computed key `globalThis[k]` → skipped undefined, 1 replacement; comma-wrapped `Object.assign((0, Deno.build), …)` → skipped undefined, 1 replacement. Suite after: `ok | 29 passed | 0 failed`. Full dry run (`--rules fs,temp,env,process,platform --dir src --dry-run`) unchanged: `src/core/deno/monkey-patch.ts: skipped: assigns to Deno.realPathSync at line 17`, `src/core/performance/metrics.ts: skipped: assigns to Deno.readTextFileSync at line 23`, `Summary: 624 files scanned, 224 would be rewritten, 816 replacements, 18 with collisions, 2 skipped`; `git status --short src` empty. README "Open decisions with an owner" gained the PR 20b and PR 25 rows that remove the two entries later.
 - Commit "runtime-migration docs: PR 09 plan matches its execution; tracking section is tool-neutral": private details (tracker IDs, maintainer-local worktree paths, tracker instructions) left the committed docs; README progress tracking is tool-neutral. Acceptance criteria, Verification and the review checklist now match what was executed: ral suite `ok | 24 passed | 0 failed` (on Windows `23 passed | 1 ignored`), codemod suite `ok | 29 passed | 0 failed`, and the ratchet baseline check is `git diff upstream/main...HEAD -- tools/runtime-boundary-baseline.json` showing only `tests` 874 → 875. RC1 (ReadFileOptions) was closed without a code change: the codemod reports `qualified-path.ts`'s own `readTextFile`/`readTextFileSync` as collisions, no caller passes `options`, and an aborted read's `DOMException.message` is getter-only, so a `signal` in the layer would break `errorWithPath`; only the Context line records the decision. Step 3 checks: both greps printed nothing; baseline diff is the one hunk above; suites re-run after the edits: codemod `ok | 29 passed | 0 failed`, unit ral `ok | 23 passed | 0 failed | 1 ignored`.
+
+### Rebase onto main (#15218)
+
+- `git fetch upstream && git rebase upstream/main`: upstream/main was one commit ahead of the branch base (`623cd99e8`, PR 02d); 36 commits replayed. Only `dev-docs/runtime-migration/HANDOFF.md` conflicted — the status line and "How to resume" step 5, which both sides edited; resolved to "PR 02 (#15135), 02c (#15153) and 02d (#15218) are merged, 09 (the `src/deno_ral/` fs, temp and errors layer) is in progress, and the next to run are 00a, then 00b" and "PR 02, 02c and 02d are merged, 09 is in progress; …". README auto-merged.
+- First codemod run after the rebase: `FAILED | 15 passed | 14 failed`, every failure a pure line-ending diff. Replaying the branch re-checked `tools/codemod/**` out, and `core.autocrlf=true` converted the LF blobs to CRLF working-tree files (`git ls-files --eol` showed `i/lf w/crlf`), while the fixture tests compare bytes verbatim and ts-morph inserts lines as `\n`. Fixed by pinning `tools/codemod/** text eol=lf` in `.gitattributes` (commit "tools: pin the codemod test files to LF checkouts") and re-checking out: all 76 files now `i/lf w/lf`.
+- `package/dist/bin/quarto.cmd run --dev tools/check-runtime-boundary.ts`:
+
+  ```
+  scope         Deno.* calls            std-import files
+  src           992 (baseline 992)      49 (baseline 49)
+  src-js        7 (baseline 7)          0 (baseline 0)
+  resources     12 (baseline 12)        1 (baseline 1)
+  package/src   240 (baseline 240)      3 (baseline 3)
+  tests         875 (baseline 875)      176 (baseline 176)
+  OK: no count exceeds the baseline.
+  ```
+- `git diff upstream/main...HEAD -- tools/runtime-boundary-baseline.json`: one hunk, `tests` `"denoCalls": 874,` → `"denoCalls": 875,`.
+- `package/dist/bin/tools/x86_64/deno test --allow-all --import-map=src/import_map.json tools/codemod/` → `ok | 29 passed | 0 failed (705ms)`.
+- `cd src && ../package/dist/bin/tools/x86_64/deno check --unstable-kv --unstable-ffi --import-map=import_map.json quarto.ts` fails on Deno 2.7 with the import-access error recorded above; with `--allow-import` (the recorded workaround) → `Check quarto.ts`, exit 0.
+- ral suite: `run-tests.ps1 --agent unit/ral/` is rejected ("Only .ts, or .qmd, .md and .ipynb passed to smoke-all.test.ts are accepted") and `run-tests.sh` on Windows looks for `package/dist/bin/tools/deno`, which does not exist (the binary is under `tools/x86_64/`), so the five test files were passed explicitly: `cd tests && pwsh -Command '$env:QUARTO_TESTS_NO_CONFIG="true"; .\run-tests.ps1 --agent unit/ral/errors.test.ts unit/ral/fs.test.ts unit/ral/safe-exists-file.test.ts unit/ral/safe-remove-dir.test.ts unit/ral/temp.test.ts'` → `ok | 23 passed | 0 failed | 1 ignored` ("safeRemoveDirSync with symlinks" is `ignore: isWindows`).
@@ -76,17 +76,6 @@ The C-row correction of `aliasTarget()` in PR 36 is subsumed by PR 36 Decision 4
 
 Two notes for whoever continues: the B1 commit carries the maintainer's author identity by mistake (an amend and force-push was refused; the attribution line is in the message); the triage's C rows for PRs 33, 35 and 36 overlapped files that B2, B4, B7 and B8 edited, and were run after the pass landed (all three are done).
 
-## Pending on `rt/09` before its PR opens
-
-Written 2026-10-28; delete this section once both items are done.
-
-1. Rebase onto `main`: PR 02d merged as #15218 (`623cd99e8`).
-   `HANDOFF.md` conflicts on the two status lines both branches edited; resolve them to "02d (#15218) is merged, 09 is in progress".
-   After the rebase, set the README `02d` row to `merged`.
-   `tools/runtime-boundary-baseline.json` merges without a conflict; `quarto run --dev tools/check-runtime-boundary.ts` should then report `src` 992, `resources` 12, `tests` 875 and no "rose" message.
-2. Task 4 review: the committed codemod (`c5af406bc`) had four defects, each to be fixed with a fixture before Task 5.
-   Reverse mode turned a shorthand property (`{ os }`) into invalid syntax (`{ Deno.build.os }`); a type-only namespace import (`import type * as fs`) was overwritten with named imports; an existing `import { type statSync }` plus a `Deno.statSync(...)` call produced a duplicate binding; and forward mode rewrote calls on a local binding named `Deno`.
-
 ## How to resume
 
 1. Work from an up-to-date `main`: the plans are there.