emrg: the release remedy reads the title, not the body

scripts/check-issue-links.py

@@ -1399,15 +1399,29 @@ def _mention_detail(refs: Refs, number: int) -> str:
     return "; ".join(clauses)
 
 
-#: A release version an issue's own text names, e.g. `Release v0.3.46`. It is read
+#: A release version an issue's own **title** names, e.g. `Release v0.3.46`. It is read
 #: because it is the one fact an open issue carries about what finishes *it*: a PR's own
 #: title never reaches `judge_issues`, so "was the declaring PR a bump?" cannot be answered
 #: here at all, while "does this issue's completion name a published artifact?" can.
+#:
+#: The title and not the body, because the two say different things: a title says what an
+#: issue **is**, while a body says what it was **seen in**. An ordinary issue cites the
+#: release it observed its defect on, and reading that as the artifact that finishes it
+#: hands the reader a published version for work that has nothing to do with one - the
+#: reading then exits 0 and invites a close over work that is not done. Measured over the
+#: repository's last 200 issues: 34 name a version, 26 of them in the title and 8 only in
+#: the body, and three of those 8 driven through `judge_issues` each took the artifact
+#: remedy for a merge about something else. An issue naming its version only in the body
+#: therefore keeps the remedy it had.
 _RELEASE_VERSION = re.compile(r"\bv\d+\.\d+\.\d+\b")
 
 
 def release_version(*texts: str | None) -> str | None:
-    """The release version this issue's own text names, or `None` when it names none."""
+    """The release version the given text names, or `None` when it names none.
+
+    Which text is the caller's decision, not this function's: `judge_issues` passes the
+    issue's title alone, and the reason is stated on `_RELEASE_VERSION`.
+    """
     for text in texts:
         if not text:
             continue
@@ -1521,9 +1535,7 @@ def judge_issues(
                 )
                 continue
             if refs.declared_merged or refs.declared_abandoned:
-                detail = _declined_detail(
-                    refs, number, release_version(issue.get("title"), issue.get("body"))
-                )
+                detail = _declined_detail(refs, number, release_version(issue.get("title")))
             elif refs.mentioned_landed or refs.mentioned_abandoned or refs.mentioned_open:
                 detail = _mention_detail(refs, number)
             else:

tests/test_check_issue_links.py

@@ -2854,3 +2854,45 @@ def test_judge_issues_reads_the_version_off_the_issue_it_is_judging(mod) -> None
     rows = {row.number: row for row in mod.judge_issues(issues, refs, {})}
     assert "check-release-published.py v0.3.46" in rows[2781].detail
     assert "check-release-published" not in rows[2778].detail
+
+
+# --- the title declares the version, the body only mentions one --------------------
+#
+# The artifact remedy belongs where an issue **declares** the release it is about, and is
+# wrong where its body merely cites the release it was **seen in**: that version is already
+# published, so the printed reading exits 0 and invites a close over work it never
+# described. Measured over the repository's last 200 issues, 34 name a version and 8 of
+# them name it only in the body - each of those took the artifact sentence before this
+# gate, for a merge that had done something else.
+
+
+def test_a_version_in_the_body_does_not_make_an_issue_a_release_issue(mod) -> None:
+    issues = [
+        {
+            "number": 2760,
+            "title": "The actionlint step's download has no retry, so one upstream 504 reddens master for hours",
+            "body": "master has failed its Test workflow since a85b1ddd; v0.3.41 stepped on the download.",
+            "created_at": "2026-11-08T10:00:00Z",
+        }
+    ]
+    refs = {2760: mod.Refs(declared_merged={2759})}
+    rows = {row.number: row for row in mod.judge_issues(issues, refs, {})}
+    assert "check-release-published" not in rows[2760].detail
+    assert "either close this issue with the reading that says the work is done" in rows[2760].detail
+
+
+def test_the_title_is_what_declares_the_version_and_the_body_is_not(mod) -> None:
+    body = "master has failed its Test workflow since a85b1ddd; v0.3.41 stepped on the download."
+    issues = [
+        {"number": 2781, "title": "Release v0.3.46", "body": body, "created_at": "2026-11-09T17:35:22Z"},
+        {
+            "number": 2760,
+            "title": "The actionlint step's download has no retry, so one upstream 504 reddens master for hours",
+            "body": body,
+            "created_at": "2026-11-08T10:00:00Z",
+        },
+    ]
+    refs = {2781: mod.Refs(declared_merged={2782}), 2760: mod.Refs(declared_merged={2759})}
+    rows = {row.number: row for row in mod.judge_issues(issues, refs, {})}
+    assert "check-release-published.py v0.3.46" in rows[2781].detail
+    assert "check-release-published" not in rows[2760].detail