A gate deciding by machine, and a lookup that never found its subject: two rows with their controls

dag/gunbc/recurring_failure_mode/absent_reads_identically_to_never_looked.dag

@@ -19,6 +19,14 @@ data absent_reads_identically_to_never_looked: RecurringFailureMode = RecurringF
 
     "THE REPAIR IS THAT THE INSTRUMENT MUST NOT BE ABLE TO RETURN A VALUE IT CANNOT SHOW IT RESOLVED. A positive control that must hit, or a denominator it must report -- a path read asserts the source is non-empty before its emptiness is allowed to mean anything, a windowed read asserts the window contains a known landmark, a directory search asserts the directory exists. A retry does not repair it: the same wrong path, window or directory returns the same empty result and agrees with itself. ",
 
+    "FOURTH SPECIMEN, AND IT MOVES THE CLASS FROM OUR OWN ADDRESSING TO AN EXTERNAL TOOL'S (gunbc#10311, 2026-09-04). `gh run list --repo gunb-ai/gunbc --commit <11-character sha>` returned `[]` with EXIT 0; the same query with the full forty-character sha returned TWO runs -- a queued workflow_dispatch and a completed pull_request. Same head, same tool, same minute, reproduced independently by a second session on a different head within the hour. The three earlier specimens are all the caller mis-addressing its own read; this one is a tool ACCEPTING an identifier, failing to resolve it, and rendering that as a filter matching nothing. It is the same class because the same capability retires all four -- an instrument may not return a value it cannot show it resolved -- but it widens the population from reads we author to every external query keyed on an identifier. ",
+
+    "**THE INPUT THAT BREAKS THE FOURTH SPECIMEN IS THE DEFAULT FORM IN WHICH WE HAND EACH OTHER SUBJECTS, WHICH IS WHAT MAKES IT A FLEET HAZARD RATHER THAN A TOOL QUIRK.** Every status message, review comment and hand-off in this repository quotes ten- and eleven-character shas, because that is what git prints and what prose wants. So the abbreviated form is not a careless variant a disciplined reader avoids -- it is the form the subject ARRIVES in. And the question it was asked was whether a healed head had been picked up for revalidation, which is exactly the question whose reassuring answer ends the investigation: the honest report would have been STRANDED, and the truth was merely QUEUED. ",
+
+    "TRIGGER CONTRIBUTED BY THE FOURTH SPECIMEN, AND IT IS NARROWER AND MORE ACTUATABLE THAN A POSITIVE CONTROL: WHERE THE SUBJECT IS AN IDENTIFIER, CANONICALIZE IT BEFORE ASKING. For a git object that means forty characters, and it means treating ANY sha that arrived in prose as unresolved until expanded. That is decidable on our side even though the tool's behaviour is not ours to change -- DESIGN's outside-the-modeled-guarantee boundary bounds the CEILING at mechanically preventable for our own callers, and no higher. ",
+
+    "BOUNDED AGAINST `empty_capture_read_as_clean_result`, which is the nearest neighbour and is NOT this: there the instrument DID refuse and said so, and the refusal merely travelled on a channel the reader did not keep, so the information exists and the repair is to keep the other stream. Here NO refusal exists anywhere, so there is no discarded diagnostic to recover. ",
+
     "Distinct from `denominator_moved_between_measurement_and_comparison`, where the instrument reached a real subject that had changed underneath it; here the subject never moved and was never reached. Both are cases of a number that describes the measuring rather than the measured, which is why neither is caught by re-running.)",
   ],