diff --git a/scripts/tests/test_hook_fire_log.py b/scripts/tests/test_hook_fire_log.py
index 87b976333..8aa8a1c4f 100644
--- a/scripts/tests/test_hook_fire_log.py
+++ b/scripts/tests/test_hook_fire_log.py
@@ -151,7 +151,13 @@ def instrumentation_faults(text: str, name: str) -> list[str]:
     # `#` is a comment in every shell. `pretooluse-merge-consent.sh` legitimately discusses this
     # variable in prose. Every other line mentioning it must be one of the two canonical lines, in
     # order. A heredoc carrying the token would fault; no hook has one, and refusing is the safe
-    # direction — this arm's errors are refusals, never acceptances.
+    # direction for the lines this arm COMPARES — but not for the ones it SELECTS. `mentions` is
+    # chosen by the literal `ETV_HOOK_FIRE_LIB`, so a writer of that variable which never spells the
+    # literal — one composed at runtime and `eval`ed, say — is outside the compared set altogether,
+    # and byte-identity WITHIN a selection is not byte-identity over the file
+    # (`testing.guard-pins-the-artifact-not-a-shape`, which states this pin's reach because the pin
+    # itself does not). So this arm's errors are not all refusals, and the `split("\n")` residual
+    # below is the second acceptance of the same kind.
     # `split("\n")`, not `splitlines()`: the latter also breaks on \x0b, \x0c, \x1c-\x1e, \x85 and the
     # Unicode separators, none of which ends a line for the shell. The divergence runs BOTH ways —
     # a canonical assignment followed by `\x0c` and garbage on one physical line parses clean under