test(of): pin the Nativity Vigil at entry 3, not entry 9

Fix round 1 (coordinator review). The Nativity Vigil (24 December,
Sollemnitas) bands at entry 3 (30) rather than entry 9 (90) as a direct
consequence of this task's ruling, but nothing pinned it -- a future
branch-reorder (e.g. inserting an entry between 3 and 9, or loosening
entry 9's date guard) could silently revert it to 90 with the suite
green.

Adds one row to part2_cases with a comment recording: the Tabula's own
entry 9 textually covers 24 December as a late-Advent feria, yet Phase
1 gives that civil day the of-nativity-vigil identity at Sollemnitas
because the Vigil Mass is celebrated that evening -- the same
civil-day-granularity shape Holy Thursday already settles (entry 2,
not entry 1); Phase 2 classifies Phase 1's emitted identity rather
than overriding it. Unobservable on the universal calendar today (no
competing candidate on 24 December); live under a proper/diocesan
overlay, where a proper solemnity (entry 4, band 40) loses to the
Vigil at 30 but would beat it at the textually-arguable 90.

No production code change -- test-only, per the coordinator's ruling
to pin current behaviour rather than alter it.

1 files changed, 25 insertions, 0 deletions

diff --git a/test/test_precedence_of.ml b/test/test_precedence_of.ml
index a06fb25..0d2db63 100644
--- a/test/test_precedence_of.ml
+++ b/test/test_precedence_of.ml
@@ -135,6 +135,31 @@ let part2_cases =
       ctx ~iso:"2026-06-04" ~season:Vocab_of.Ordinary_time,
       mk ~slug:"of-corpus-christi" ~rank:Vocab_of.Sollemnitas ~origin:Precedence.Temporal (),
       30 );
+    (* Fix round 1 (coordinator review, 2026-08-25). (a) THE TENSION: the
+       Tabula's own entry 9 reads "Feriae Adventus a die 17 ad 24 decembris
+       INCLUSIVE" -- 24 December is textually a late-Advent feria -- yet
+       Phase 1's temporal_of.ml gives that civil day the identity
+       [of-nativity-vigil] at rank Sollemnitas, because the Vigil Mass is
+       celebrated that evening; at colitur's one-office-per-civil-day
+       granularity only one identity can hold the day. (b) THE RESOLUTION:
+       Phase 2 classifies whatever identity Phase 1 already chose to emit, it
+       does not reach back and override that choice -- the same shape, and
+       the same settlement, as Holy Thursday (entry 2, not entry 1) above.
+       So this candidate's rank (Sollemnitas) and layer (temporal, hence
+       [not (is_proper layer)]) correctly route it to entry 3, band 30, not
+       entry 9's 90. (c) OBSERVABILITY: on the universal calendar nothing
+       else competes for 24 December, so this value can never be OBSERVED
+       today -- it is unobservable, not untested. It becomes live the moment
+       a proper/diocesan overlay places a proper solemnity (entry 4, band 40)
+       on 24 December: at 30 the Vigil beats it, but at the textually-arguable
+       90 the proper solemnity would beat the Vigil instead. That reversal is
+       exactly what this row guards against a silent branch-reorder. *)
+    ( "entry 3, not 9: the Nativity Vigil is a solemnity, though 24 December \
+       is also a late-Advent feria",
+      ctx ~iso:"2026-12-24" ~season:Vocab_of.Advent,
+      mk ~slug:"of-nativity-vigil" ~rank:Vocab_of.Sollemnitas
+        ~origin:Precedence.Temporal (),
+      30 );
     ( "entry 5: the Baptism of the Lord is a feast OF THE LORD",
       ctx ~iso:"2026-01-11" ~season:Vocab_of.Christmas,
       mk ~subject:Subject.Lord ~slug:"of-baptism-of-the-lord" ~rank:Vocab_of.Festum