diff --git a/docs/doctoring/opencode-peer-check-reevaluation.md b/docs/doctoring/opencode-peer-check-reevaluation.md
new file mode 100644
index 0000000000..56509accb5
--- /dev/null
+++ b/docs/doctoring/opencode-peer-check-reevaluation.md
@@ -0,0 +1,56 @@
+# OpenCode peer-check fallback reevaluation
+
+Status: Proposed. Owner issue: #2125.
+
+Protected main `fb17ef556f94f673234aa557254ae52779e9a7b0` accepted a
+current-head CHANGES_REQUESTED review carrying its canonical model-unavailable
+fallback marker as a substantive receipt. The required caller consequently
+skipped another review after peer CodeQL checks recovered (#2113,
+job 103581933696).
+
+The receipt helper matches the complete canonical failed-peer-check payload:
+fixed producer prose, the current head SHA, failed-check rows, and an optional
+generated Mermaid evidence map. Extra prose, mixed findings, and unknown formats
+remain formal blockers; a heading or fallback marker cannot remove a product
+finding. A diagram heading alone cannot hide additional prose. A newer fallback blocks reuse of an older same-head
+receipt; a later substantive product finding still deduplicates normally.
+Approval eligibility and downstream gates are unchanged. This does not approve
+any PR or establish model availability. PR #1706 edits a separate verdict lookup
+in the same caller; this repair changes only the shared receipt helper.
+
+Regression: six fallback-marker cases failed before the change; a substantive
+finding case already passed. The receipt, live-draft and required-verdict tests
+exercise the shared boundary and existing caller behavior. Actual fresh receiver
+execution and a formal current-head review remain deployment acceptance gates.
+
+## Concurrent owner integration
+
+Ordinary merge preserves both aa6150c and 96704bc7 histories. The remote
+substantive-finding repair's original mixed-review case is retained as an
+additional regression. Its broad heading regex is superseded by the exact
+producer envelope because unknown headings must remain blockers, not silently
+qualify for reevaluation. No approval criterion is relaxed.
+
+### Historical remote-branch validation (not merged-head proof)
+
+Regression evidence is bound to exact commits. Commit
+`9cd835df682b74ead1e5306b92280da941e45040` is RED because a mixed fallback
+and authorization finding returns no receipt. Commits
+`7b528279720af353ba91fbff5d43384ad466bb8b` and
+`f5be0fb8f21d9fad64c8576b5e980d5eeb9a1b1e` introduce and correct the
+minimal finding-heading classifier. On the corrected source blob
+`310f05d956382d4ba525907ccba64c633c1c2a6c` and test blob
+`05955d2e2181aefe957f239d8aae8012a0ad04d9`, the receipt,
+live-Draft, required-verdict, and coverage-publication suites report 86 passed;
+Ruff and compileall are green. Hosted exact-head checks, independent review, and
+a fresh #2113 receiver/formal review remain deployment acceptance gates.
+
+## Unstructured finding regression
+
+The db8058e9 test exposes a product finding appended as ordinary prose beneath
+the canonical heading. The former heading-only classifier discarded it. The
+replacement matches the full producer envelope and fails closed on added prose,
+including text after the diagram. Tests execute the actual producer printf block
+and the graph emitter rather than reproducing shortened synthetic envelopes.
+Receipt-module validation: 25 passed, 100% statements and branches; hosted
+review and deployment evidence remain outstanding.
diff --git a/scripts/ci/opencode_review_receipt_gate.py b/scripts/ci/opencode_review_receipt_gate.py
index 4dcb24af88..279634a40e 100644
--- a/scripts/ci/opencode_review_receipt_gate.py
+++ b/scripts/ci/opencode_review_receipt_gate.py
@@ -110,6 +110,36 @@ def is_mention_or_malformed(body: str) -> bool:
     return not any(marker in stripped for marker in PRODUCT_MARKERS)
 
 
+def is_peer_check_only_fallback(body: str, head_sha: str) -> bool:
+    """Match the complete producer payload; retain unknown prose as a blocker."""
+    prefix = (
+        "## Pull request overview\n\n"
+        "OpenCode could not approve from deterministic current-head evidence "
+        "because GitHub Checks have failed.\n\n"
+        "## Findings\n\n"
+        "### 1. HIGH Current-head GitHub Checks - Fix failed required checks before approval\n"
+        f"- Problem: Failed same-head checks remain for `{head_sha}`.\n"
+        "- Root cause: The model-unavailable evidence fallback is allowed only "
+        "when peer GitHub Checks are complete and clean.\n"
+        "- Fix: Read and fix the failed check logs below, then rerun the current-head checks.\n"
+        "- Regression test: Keep the model-unavailable fallback gated on an empty "
+        "failed-check rollup.\n\nFailed checks:\n"
+    )
+    check_line = (
+        r"- [^\n\r]+: (?i:FAILURE|ERROR|TIMED_OUT|ACTION_REQUIRED|CANCELLED|STARTUP_FAILURE)"
+        r"(?: \(https://[^\s()]+\))?"
+    )
+    graph = (
+        r"\n\n## Changed-File Evidence Map\n\n```mermaid\n"
+        r"(?:flowchart LR|classDiagram|sequenceDiagram)\n(?:  [^\n`]+\n)+```"
+    )
+    return re.fullmatch(
+        re.escape(prefix) + check_line + r"(?:\n" + check_line + r")*"
+        + r"(?:" + graph + r")?\n?",
+        body,
+    ) is not None
+
+
 def is_formal_receipt(
     review: Mapping[str, Any],
     head_sha: str,
@@ -134,6 +164,10 @@ def is_formal_receipt(
         marker in body.casefold() for marker in FALLBACK_APPROVAL_MARKERS
     ):
         return False, "fallback approval is not a substantive formal review"
+    if state == "CHANGES_REQUESTED" and is_peer_check_only_fallback(body, head_sha) and any(
+        marker in body.casefold() for marker in FALLBACK_APPROVAL_MARKERS
+    ):
+        return False, "fallback changes request requires fresh substantive review"
     if is_draft and state == "APPROVED":
         return False, "draft must never receive bot APPROVE"
     return True, "current-head formal review"
@@ -161,7 +195,7 @@ def evaluate_receipts(
             return review, reason
         if "never receive bot APPROVE" in reason:
             return None, reason
-        if "fallback approval" in reason:
+        if "fallback approval" in reason or "fallback changes request" in reason:
             return None, reason
         if reason.startswith("stale"):
             stale_hits += 1
diff --git a/tests/test_opencode_review_receipt_gate.py b/tests/test_opencode_review_receipt_gate.py
index c971e2128a..099154bd35 100644
--- a/tests/test_opencode_review_receipt_gate.py
+++ b/tests/test_opencode_review_receipt_gate.py
@@ -342,3 +342,162 @@ def unexpected_run(args, **kwargs):
         )(),
     )
     assert receipt.load_reviews("-")[0]["commit_id"] == receipt.AFIPC_230_HEAD
+
+
+def canonical_peer_fallback(head):
+    """Execute the producer's printf block with a realistic failed-check row."""
+    import subprocess
+    source = Path(".github/workflows/opencode-review-dispatch.yml").read_text()
+    # Locate the unique canonical producer directly; other fallbacks stay independent.
+    marker = source.index("printf 'OpenCode could not approve from deterministic current-head evidence because GitHub Checks have failed.")
+    start = source.rfind("                printf '## Pull request overview", 0, marker)
+    end = source.index('                cat "$failed_checks_file"', marker)
+    program = source[start:end] + "printf '%s\\n' '- CodeQL PR/CodeQL compatibility analysis (python): FAILURE (https://github.com/ContextualWisdomLab/.github/actions/runs/1)'"
+    return subprocess.check_output(["bash", "-c", program], env={"HEAD_SHA": head}, text=True)
+
+
+def test_fallback_changes_request_requires_fresh_review():
+    """The complete producer envelope triggers review without reusing approval."""
+    head = receipt.AFIPC_230_HEAD
+    fallback = review(commit=head, body=canonical_peer_fallback(head))
+    older_approval = review(commit=head, state="APPROVED")
+    found, reason = receipt.evaluate_receipts([older_approval, fallback], head)
+    assert found is None
+    assert "fallback" in reason
+
+
+def test_substantive_changes_request_still_deduplicates():
+    """Actual product findings remain a receipt even after a prior fallback."""
+    head = receipt.AFIPC_230_HEAD
+    fallback = review(commit=head, body="## Pull request overview\nmodel-unavailable evidence fallback")
+    substantive = review(commit=head, body=(
+        "## Pull request overview\n## Findings\n"
+        "### 1. HIGH Missing authorization\n"
+        "The changed endpoint allows anonymous writes.\n"
+    ))
+    found, _ = receipt.evaluate_receipts([fallback, substantive], head)
+    assert found == substantive
+
+
+
+def test_fallback_marker_does_not_hide_substantive_finding() -> None:
+    """A fallback marker cannot suppress a real product blocker in the same review."""
+    head = receipt.AFIPC_230_HEAD
+    mixed = review(
+        commit=head,
+        body=(
+            "## Pull request overview\n"
+            "model-unavailable evidence fallback\n"
+            "OpenCode could not approve from deterministic current-head evidence because GitHub Checks have failed.\n## Findings\n"
+            "### 1. HIGH Current-head GitHub Checks - Fix failed required checks before approval\n"
+            "### 2. HIGH Missing authorization\n"
+            "The changed endpoint allows anonymous writes.\n"
+        ),
+    )
+    found, reason = receipt.evaluate_receipts([mixed], head)
+    assert found == mixed
+    assert reason == "current-head formal review"
+
+
+@pytest.mark.parametrize("suffix", [
+    "\n### 2. HIGH Missing authorization\nAnonymous writes are allowed.\n",
+    "\n### Unexpected finding shape\n",
+    "\n#### Missing authorization\n",
+    "\n###Missing authorization\n",
+])
+def test_unknown_or_mixed_finding_is_retained(suffix):
+    """Only an exact peer-check-only finding list may trigger reevaluation."""
+    head = receipt.AFIPC_230_HEAD
+    body = ("## Pull request overview\nmodel-unavailable evidence fallback\n"
+            "OpenCode could not approve from deterministic current-head evidence because GitHub Checks have failed.\n## Findings\n"
+            "### 1. HIGH Current-head GitHub Checks - Fix failed required checks before approval\n")
+    candidate = review(commit=head, body=body + suffix)
+    assert receipt.evaluate_receipts([candidate], head)[0] == candidate
+
+
+def test_evidence_map_headings_are_not_product_findings():
+    """Only the generated fenced diagram may follow the complete finding."""
+    from scripts.ci.opencode_review_surfaces import emit_mermaid
+    head = receipt.AFIPC_230_HEAD
+    body = canonical_peer_fallback(head) + "\n## Changed-File Evidence Map\n\n" + emit_mermaid([])
+    assert receipt.evaluate_receipts([review(commit=head, body=body)], head)[0] is None
+    for extra in ("\nThe endpoint allows anonymous writes.", "\n## Findings\nOther finding"):
+        candidate = review(commit=head, body=body + extra)
+        assert receipt.evaluate_receipts([candidate], head)[0] == candidate
+
+
+def test_unknown_fallback_format_is_retained():
+    """A fallback marker alone cannot classify an unknown review as peer-only."""
+    head = receipt.AFIPC_230_HEAD
+    candidate = review(commit=head, body="## Pull request overview\nmodel-unavailable evidence fallback")
+    assert receipt.evaluate_receipts([candidate], head)[0] == candidate
+
+
+def test_peer_fallback_literals_remain_bound_to_canonical_producer():
+    """Producer wording drift requires an explicit receipt-contract update."""
+    source = Path(".github/workflows/opencode-review-dispatch.yml").read_text()
+    assert "OpenCode could not approve from deterministic current-head evidence because GitHub Checks have failed." in source
+    assert "### 1. HIGH Current-head GitHub Checks - Fix failed required checks before approval" in source
+
+
+def test_mixed_finding_without_canonical_overview_remains_blocking():
+    """Preserve the concurrent owner's original mixed-review counterexample."""
+    head = receipt.AFIPC_230_HEAD
+    candidate = review(commit=head, body=(
+        "## Pull request overview\nmodel-unavailable evidence fallback\n"
+        "## Findings\n"
+        "### 1. HIGH Current-head GitHub Checks - Fix failed required checks before approval\n"
+        "### 2. HIGH Missing authorization\nThe changed endpoint allows anonymous writes.\n"
+    ))
+    assert receipt.evaluate_receipts([candidate], head)[0] == candidate
+
+
+@pytest.mark.parametrize("sections", ["", "## Findings\n### 1. HIGH Missing authorization\n## Findings\n### 2. HIGH Missing authorization\n"])
+def test_missing_or_duplicate_findings_remain_blocking(sections):
+    """Ambiguous section structure must never discard an active change request."""
+    head = receipt.AFIPC_230_HEAD
+    candidate = review(commit=head, body=(
+        "## Pull request overview\nmodel-unavailable evidence fallback\n"
+        "OpenCode could not approve from deterministic current-head evidence "
+        "because GitHub Checks have failed.\n" + sections
+    ))
+    assert receipt.evaluate_receipts([candidate], head)[0] == candidate
+
+
+def test_peer_fallback_with_unstructured_product_finding_remains_blocking() -> None:
+    """Substantive prose cannot hide under the canonical peer-check heading."""
+    head = receipt.AFIPC_230_HEAD
+    candidate = review(
+        commit=head,
+        body=(
+            "## Pull request overview\n\n"
+            "OpenCode could not approve from deterministic current-head evidence "
+            "because GitHub Checks have failed.\n\n"
+            "model-unavailable evidence fallback\n\n"
+            "## Findings\n\n"
+            "### 1. HIGH Current-head GitHub Checks - Fix failed required checks before approval\n"
+            f"- Problem: Failed same-head checks remain for `{head}`.\n"
+            "- Root cause: The model-unavailable evidence fallback is allowed only "
+            "when peer GitHub Checks are complete and clean.\n"
+            "- Fix: Read and fix the failed check logs below, then rerun the current-head checks.\n"
+            "- Regression test: Keep the model-unavailable fallback gated on an empty "
+            "failed-check rollup.\n\n"
+            "Failed checks:\n"
+            "- CodeQL PR/CodeQL compatibility analysis (python): FAILURE "
+            "(https://github.com/ContextualWisdomLab/.github/actions/runs/1)\n\n"
+            "The changed endpoint also allows anonymous writes.\n"
+        ),
+    )
+
+    found, _ = receipt.evaluate_receipts([candidate], head)
+
+    assert found == candidate
+
+
+@pytest.mark.parametrize("extra", ["The endpoint allows anonymous writes.\n", "- Missing authorization\n", "## Other finding\n"])
+def test_canonical_payload_with_extra_prose_is_retained(extra):
+    """Unknown prose anywhere in an otherwise valid payload remains blocking."""
+    head = receipt.AFIPC_230_HEAD
+    for body in (extra + canonical_peer_fallback(head), canonical_peer_fallback(head) + extra):
+        candidate = review(commit=head, body=body)
+        assert receipt.evaluate_receipts([candidate], head)[0] == candidate