diff --git a/dag/gunbc/recurring_failure_mode/witness_module_absent_from_the_executed_set.dag b/dag/gunbc/recurring_failure_mode/witness_module_absent_from_the_executed_set.dag
index 5b6f7e77b75..b20440b506f 100644
--- a/dag/gunbc/recurring_failure_mode/witness_module_absent_from_the_executed_set.dag
+++ b/dag/gunbc/recurring_failure_mode/witness_module_absent_from_the_executed_set.dag
@@ -27,6 +27,8 @@ data witness_module_absent_from_the_executed_set: RecurringFailureMode = Recurri
 
     "SO THE ADMISSION MECHANISM MATTERS, AND IT IS NOT ALWAYS AN OMISSION. The instances above are modules nobody scheduled -- a gap. This one was scheduled OUT, deliberately, by a path-policy rule that admits an entry as deferred: the roster that records it is itself the evidence, and it is honest about what it is. The class is the same because the consequence is identical (no executing consumer), but the DETECTION differs: a gap is found by joining modules-on-disk to modules-scheduled, while a deferral is already written down and needs only to be read as what it says. A deferral roster is therefore a list of modules whose reds nobody will see, and it can be consulted directly rather than derived.",
 
+    "SPECIMEN, 2026-10-07, the same absence HIDING A STANDING RED, and the strongest form of the second harm. test.claim.self_host_emitted_call_target_realization_witness_test::w_depth_one_shared_field_refutation_stays_shape_guarded FAILS against main at the IMMUTABLE sha 1728bcc0ab4e7d8285359f161612d4c1226812f2 (Gate parse_acceptance_census: accepted arm on the smallest module, census phase 3, #13415) -- claim_batch claim-run at that sha, exit=1, FAIL=1, BuildBuddy remote dispatch completed 2026-10-07T08:27:47.537Z, discovering session vivid-lynx-377 job j36fc601a, command and output retained in the session transcript -- and every unrepaired descendant of that sha inherits the red. No STANDING gate executed it: the ordinary prepared subject excludes modules outside the required-gate closure, and changed-witness selection replaces that disposition only when a diff selects the identity -- which is exactly how the red surfaced: an unrelated PR touched the entry's sibling probes, inherited three defunct-import reds from the #13388 re-homing, and ran the whole set. The failing witness synthesizes a std-types-only probe (a shared coproduct matched at depth one inside a record pattern) and imports neither v2.std.optional nor any renamed spelling; the red predates every edit on the discovering branch. THE HONEST CEILING IS RUNG 2: a red witness on main stays writable -- a scheduled full-corpus lane would only expose and block it -- so rung 3 is not on offer. THE HONEST NEXT-RUNG TRIGGER is that lane IMPLEMENTED AND ENROLLED, not a sign-off: a standing claim run over a DEFINED CLAIM-ROOT POPULATION that refuses, or invokes a declared remedy, on every red or non-verdict, DEMONSTRATED FAILING on a planted or standing red. Manual attribution -- running the entry on main's own tree before blaming the discovering PR -- is how this specimen was found, and it is a discipline, not a rung.",
+
   ],
 
   evidence: [],