fix: main shipped unresolved conflict markers; guard against a repeat

`main` carried literal `<<<<<<< HEAD` / `=======` / `>>>>>>>` markers in
memory.md and deployed them to the live Space. Introduced by b676112, a
cherry-pick of the hermetic-suite commit (dbaeaa3) that hit an append-only
collision, was committed unresolved, and was pushed.

Two things made it hard to spot:

- b676112's message describes different work ("verify: routing leg passes;
dispatch blocked by private-Space 404") while its diff is the hermetic
suite. Searching the log for the socket guard does not find it.
- The code half landed correctly. conftest's guard and test_hermetic.py are
complete and right on main; only the prose merge broke. Tests were green
the entire time, because nothing tests prose.

Resolution: both colliding entries kept in chronological order. An
append-only log has no real conflict, only an ordering question.

Also corrected a statement main was making about itself. That cherry-pick
carried a TODO item reading "*This branch's* RESERVED.md is stale — it stops
at 0005". True where it was written; false on main, whose ledger correctly
shows 0006 + 0007 published and next free 0008. Rewritten to keep the lesson
(read `git show main:docs/adr/RESERVED.md`, never your working copy) without
the false claim.

NOT carried over from that branch's follow-up (52bb514): its narrative
records a rebase onto main, but main received a merge (80b5e01); and it flags
pancreas_cptac_gdc / pdac_msk_2024 as missing source_import_date, which is
not true here — all six artifacts on main carry it, verified, because the
merge re-curated all six rather than trusting a textual JSON merge. Importing
it would have written a history that did not happen and a closed backlog item.

New: test_no_unresolved_conflict_markers_are_committed, over git ls-files.
Prose is the only part of this repo nothing verifies — code has tests,
artifacts have `src.curate --check`, licences have the review table. Verified
against the real failure: restoring b676112's memory.md fails the test on all
three marker lines.

Tests: 167 pass.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

TODO.md

@@ -229,11 +229,14 @@ the design doc's M0–M4 is the intent behind it — see that doc's §0).
 - [ ] Optional: promote the stub-vs-restructure choice to an ADR if it proves contested.
       **Next free is 0008** — `main` has 0006 (variant×subtype join) *and* 0007 (re-curation
       cadence) published. Claim in `RESERVED.md` on `main`, alone, first.
-- [ ] ⚠️ **This branch's `docs/adr/RESERVED.md` is stale — it stops at 0005 and says "next free
-      0006", both of which are wrong.** It is the per-lane-copy failure the ledger's own postmortem
-      describes: the shared-row conflict cannot fire between branches, so `ls docs/adr/` and the
-      local ledger will both hand you a burned number. Check with
-      `git show main:docs/adr/RESERVED.md` before claiming, and rebase this branch.
+- [x] ⚠️ **ADR-ledger near-miss (the fourth), resolved.** A working branch's
+      `docs/adr/RESERVED.md` stopped at 0005 and said "next free 0006" while `main` had already
+      burned 0006 *and* 0007. `main`'s ledger is correct — 0006 + 0007 published, **next free
+      0008** — so nothing was mis-numbered. Kept because the near-miss is the lesson: this is the
+      per-lane-copy failure the ledger's own postmortem describes, and the shared-row conflict
+      cannot fire between branches, so `ls docs/adr/` and your local ledger will both hand you a
+      burned number. **Read `git show main:docs/adr/RESERVED.md` before claiming, never your
+      working copy.**
 
 ## Open questions
 

memory.md

@@ -839,7 +839,6 @@ by running the suite with `socket.connect` blocked outright — **111 tests pass
 all** (the three UI test modules still touch it, but that is Gradio's mocked-OAuth HF call at
 import in a local dev env, not our code and not present on a Space).
 
-<<<<<<< HEAD
 
 ---
 
@@ -866,7 +865,6 @@ misled again. One contract, with instructions when you miss it.
 Ordering is gate-first: an unauthenticated caller gets `denied` and learns nothing about our
 inputs — pinned by `test_argument_errors_are_reported_only_after_the_auth_gate`. 132 tests (7 new,
 alongside the denial-diagnosis guards they mirror).
-=======
 ---
 
 ## 2026-07-28 — Hermetic is now *enforced*, and the whole suite is hermetic
@@ -940,7 +938,6 @@ reading `SHOWCASE_STATUS.md` before updating it — the same habit that caught c
 work — `gradio[oauth]==6.18.0` (the `oauth` extra is required or the UI modules fail to import at
 all), plus the rest of `requirements.txt` and pytest. Previously the memory said to borrow
 `DecoupleRpy_Agent/.venv`.
->>>>>>> dbaeaa3 (tests: enforce hermeticity with a socket guard; last 3 network-touching modules closed)
 
 ---
 
@@ -974,3 +971,44 @@ credential operation for a human. Worth noting the order of failure for whoever
 `HF_TOKEN` unset the symptom is a 404 (invisible private repo); once set, the next thing exercised
 is the allow-list — and the identity it resolves to must be on `ALLOWED_IDS` here
 (`anne-voigt,cpelz741` today).
+
+---
+
+## 2026-07-28 — Repaired `main`: conflict markers were committed and deployed
+
+`main` carried **literal `<<<<<<< HEAD` / `=======` / `>>>>>>>` markers in `memory.md`**, and they
+went out to the live Space. Introduced by `b676112`, which was a cherry-pick of the hermetic-suite
+commit (`dbaeaa3`) that hit an append-only collision in `memory.md`, was committed unresolved, and
+was pushed. Two further things about that commit are worth knowing, because both make it hard to
+find again:
+
+- **Its message describes different work.** It reads "verify: routing leg passes; dispatch blocked
+  by private-Space 404", but its diff is the hermetic suite — `tests/conftest.py` (+109),
+  `tests/test_hermetic.py` (+69), plus TODO/memory. Searching the log for the socket guard will
+  not find it.
+- **The code half landed fine.** The socket guard and `test_hermetic.py` are correct and complete
+  on `main`; only the prose merge was broken. So this was cosmetically invisible — tests pass with
+  conflict markers sitting in `memory.md`, because nothing tests prose.
+
+**Repair:** kept both colliding entries (machine-endpoint argument errors, and hermetic
+enforcement) in chronological order — an append-only log has no genuine conflict, only an ordering
+question.
+
+**Also corrected a statement `main` was making about itself.** That cherry-pick carried over a
+TODO item reading "*This branch's* `RESERVED.md` is stale — it stops at 0005". True of the branch
+it was written on; **false on `main`**, whose ledger correctly shows 0006 + 0007 published and next
+free 0008. Rewritten to keep the lesson (check `git show main:docs/adr/RESERVED.md`, never your
+working copy — the per-lane-copy failure means the shared-row conflict cannot fire) without the
+false claim.
+
+**Deliberately NOT carried over** from that branch's follow-up commit (`52bb514`): its narrative
+records a *rebase* onto main, but `main` received a **merge** (`80b5e01`) instead, and it flags a
+gap — `pancreas_cptac_gdc` / `pdac_msk_2024` missing `source_import_date` — that **does not exist
+here**. All six artifacts on `main` carry it (verified), because the merge re-curated all six
+rather than trusting git's textual merge of the JSON. Importing that entry would have written a
+history that did not happen and a backlog item already closed.
+
+**The generalizable bit:** prose is the only part of this repo nothing verifies. Code has tests,
+artifacts have `--check`, licences have the review table — `memory.md` and `TODO.md` have nothing,
+so a broken merge in them survives a fully green suite and ships. Worth a lint step (grep the tree
+for conflict markers) if it happens twice.

tests/test_hermetic.py

@@ -67,3 +67,45 @@ def test_ui_modules_import_without_network(module):
     suite is supposed to have stopped having. Importing them here proves the seam still holds.
     """
     __import__(module)
+
+
+# --------------------------------------------------------------------------- #
+# Unresolved conflict markers — the one class of breakage nothing else catches
+# --------------------------------------------------------------------------- #
+def test_no_unresolved_conflict_markers_are_committed():
+    """No tracked file may contain a merge-conflict marker.
+
+    This exists because it happened: `main` shipped `memory.md` with literal `<<<<<<< HEAD`
+    markers to the live Space, and the suite stayed green the whole time. Code has tests,
+    artifacts have `src.curate --check`, licences have the review table — **prose has nothing**,
+    so a botched merge in `memory.md` or `TODO.md` is invisible to everything except a reader.
+
+    Checking tracked files via git (rather than walking the tree) keeps the venv, caches and
+    curated scratch out of it, and means a file is covered the moment it is added.
+    """
+    import subprocess
+    from pathlib import Path
+
+    root = Path(__file__).resolve().parents[1]
+    tracked = subprocess.run(
+        ["git", "-C", str(root), "ls-files", "-z"],
+        capture_output=True, text=True, check=True,
+    ).stdout.split("\0")
+
+    # Split so this file's own strings are not a match when it scans itself.
+    markers = ("<" * 7, "=" * 7, ">" * 7)
+    offenders = []
+    for name in filter(None, tracked):
+        path = root / name
+        if not path.is_file() or path.suffix in {".png", ".jpg", ".ico"}:
+            continue
+        try:
+            lines = path.read_text(encoding="utf-8").splitlines()
+        except (UnicodeDecodeError, OSError):
+            continue  # binary or unreadable — not prose, not our concern
+        for n, line in enumerate(lines, 1):
+            # A marker is only a marker at the START of a line, which is what git writes.
+            if any(line.startswith(m) for m in markers) and path.name != Path(__file__).name:
+                offenders.append(f"{name}:{n}: {line[:60]}")
+
+    assert not offenders, "unresolved conflict markers committed:\n" + "\n".join(offenders)