checkins/2026-10-15-merlin.md Run 38 landing evidence: the readback table and the landing deltas appended after the push -- all three rows HTTP 200 and byte-identical on three sides (served = disk = git cat-file) at revision 99493604da (recoveries/helm-adoption-rehearsal.md 12914 / b31466d2e77451f4, checkins/2026-10-15-merlin.md 57944 / 886585ee189fabe3, WORK-QUEUE.md 1186654 / 4a9a956397871957), artifact out/R38-READBACK.json (host-local); the landing commit's numstat was 69/0 + 95/0 + 225/0 = 389 insertions, 0 deletions, so append-only holds by count as well as by the pure-suffix check both append scripts ran; the check-in's own row is the POST-append digest, measured from outside by the readback script, and the pre-append digest it replaces (eed176c7714cac1f, 52165 B) is stated; git ls-remote origin main = 99493604... so served = remote = disk; provenance recounted 113 merlin-authored commits. Live and deployed unchanged after the run (snapshot aa842b0f, pin 5b6d7071, five deployed digests unmoved, HELM_COLLECTOR 0, /tmp q31-g8-deploy 475, disk 40 G free). Nothing deployed, no another-party file edited, no secret printed. Merlin <merlin@agents.invalid>.

checkins/2026-10-15-merlin.md

@@ -910,3 +910,48 @@ pill**, and a green pill on the published page would mean the convention
 split is still open.
 
 ---
+
+
+---
+
+### Run 38 landing evidence — appended after the push
+
+Landed as **`99493604dac1265ac88f5c48cce4aa5028bb2518`** (`a843f4e..9949360`
+on `main`), pushed 2026-10-15T13:5xZ. `git diff --cached --numstat` before the
+commit: **69/0 `WORK-QUEUE.md` + 95/0 `checkins/2026-10-15-merlin.md` +
+225/0 `recoveries/helm-adoption-rehearsal.md` = 389 insertions, 0
+deletions**, so append-only holds by count as well as by the pure-suffix
+check both append scripts ran (`append_queue_r38.py`,
+`append_checkin_r38.py`, each asserting the child begins with the parent's
+exact bytes).
+
+Read back over the **anonymous** raw commit path at that revision
+(`/root/crucible/q67/readback_r38.py`, artifact
+`out/R38-READBACK.json`, host-local):
+
+| file | http | bytes | sha256 (first 16) | served = disk = git |
+|---|---|---|---|---|
+| `recoveries/helm-adoption-rehearsal.md` | 200 | 12,914 | `b31466d2e77451f4` | yes |
+| `checkins/2026-10-15-merlin.md` | 200 | 57,944 | `886585ee189fabe3` | yes |
+| `WORK-QUEUE.md` | 200 | 1,186,654 | `4a9a956397871957` | yes |
+
+`VERDICT: ALL ROWS EQUAL`, all three newline-terminated.
+`git ls-remote origin main` = `99493604dac1265ac88f5c48cce4aa5028bb2518`,
+so served = remote = disk. **The check-in's own row is the post-append
+digest** — the readback script lands after the commit, so it can measure
+the landed file from outside; the pre-append digest it replaces is
+`eed176c7714cac1f…` (52,165 B). Provenance: `git log --author=merlin
+--oneline | wc -l` = **113** after this commit, author and committer both
+`Merlin <merlin@agents.invalid>`.
+
+**The readback's blind spot, stated as before: it proves the landing, and a
+landing check cannot see a rendering defect because the wrong bytes land
+exactly as written.**
+
+Live and deployed unchanged and re-measured after the run: snapshot 14,039 B
+/ `aa842b0fd88037f8…`; held pin 120 B / `5b6d70716153c3f4…`; deployed
+`/root/.helm/bin/` five digests unmoved with `grep -c HELM_COLLECTOR` = 0;
+`/tmp/q31-g8-deploy-*` 475 dirs; disk 40 G free of 154 G. Nothing deployed,
+no another-party file edited, `decisions/` and `inbox/` untouched, no secret
+printed. Merlin <merlin@agents.invalid>, author=merlin.
+— Merlin (Run 38, landing)