Strengthen the Assumption's octave against a demoted St. John Eudes

Demote St. John Eudes (Aug 19) from Duplex to Semiduplex -- too minor
a confessor to keep outranking the Assumption's octave. To make that
demotion actually cede the day, raise the Assumption's octave to
`wins: duplex` (same pattern already used by Pentecost's octave), so an
ordinary-day tie no longer automatically favors the occurring saint.

Promote St. Thomas of Canterbury (Dec 29) to Duplex so he keeps
winning against the Christmas octave stack now that ordinary-day ties
are no longer a given.

Also add a closing-day tie-break to octave-vs-octave precedence
(pickWinningOctave): when two active octaves tie in rank, the one on
its own closing day now wins the label contest, ahead of the existing
"more recently started" tie-break. Needed because Assumption's day 3
and St. Lawrence's own closing day (Aug 17) now tie at Duplex, and the
closing day should still govern that date's label.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

src/calendar/octaves.ts

@@ -115,11 +115,19 @@ export function strictestThreshold(octaves: ActiveOctave[]): FeastClass {
  *
  * - Highest effective rank (`wins`, already elevated on either octave's
  *   own closing day) wins outright.
- * - Tied rank: the more recently *started* octave wins (smaller
- *   `dayNumber` today) — the reasoning given was that day 1 of a newly
- *   started octave needs to be fully present, which is the whole point
- *   of it starting; the older octave that's already been running is
- *   commemorated instead, same as any octave that loses this comparison.
+ * - Tied rank, one of them on its own closing day: the closing day wins
+ *   — its "in Octava" day is the octave's own culminating day, so it
+ *   takes precedence over an ordinary day of an equally-ranked octave
+ *   even if that other octave started more recently (2026-08 user
+ *   instruction, concrete case: Aug 17, St. Lawrence's own closing day,
+ *   vs. the Assumption's ordinary day 3, both `duplex` once Assumption's
+ *   octave was strengthened — Lawrence's closing day wins).
+ * - Tied rank, neither (or both) on their own closing day: the more
+ *   recently *started* octave wins (smaller `dayNumber` today) — the
+ *   reasoning given was that day 1 of a newly started octave needs to be
+ *   fully present, which is the whole point of it starting; the older
+ *   octave that's already been running is commemorated instead, same as
+ *   any octave that loses this comparison.
  *
  * Undefined when no octave is active at all. Every other active octave
  * still gets commemorated regardless of which one wins here — this
@@ -187,6 +195,9 @@ export function pickWinningOctave(active: ActiveOctave[]): ActiveOctave | undefi
     if (rankCmp < 0) {
       return best;
     }
+    if (candidate.isClosingDay !== best.isClosingDay) {
+      return candidate.isClosingDay ? candidate : best;
+    }
     return candidate.dayNumber < best.dayNumber ? candidate : best;
   }, undefined);
 }

src/data/calendar/saints/assumption.yml

@@ -10,5 +10,16 @@ rank: duplex-1-classis
 common: common-of-the-bvm
 propers: "assumption"
 # Has a real octave (Aug 15-22) -- see calendar/types.ts's OctaveConfig.
+# Strengthened to `wins: duplex` (2026-08 user decision), same pattern
+# as pentecost-sunday.yml's own override -- a Semiduplex saint no
+# longer just ties (and, by the ordinary-day tie rule, wins) an
+# ordinary octave day here; it loses outright to the octave instead.
+# Concrete motivation: st-john-eudes.yml's demotion to Semiduplex only
+# actually cedes Aug 19 to the octave because of this. Side effect
+# accepted by the user: st-damasus.yml (Dec 11, Immaculate Conception's
+# octave) and st-nicholas-of-tolentino.yml (Sep 10, Nativity of the
+# BVM's octave) are unrelated octaves not touched by this change, so
+# they're unaffected -- this only strengthens Assumption's own octave.
 octave:
   enabled: true
+  wins: duplex

src/data/calendar/saints/st-john-eudes.yml

@@ -4,9 +4,16 @@
 # No [Ant 1] of his own -- benedictusCommon/minorHoursCommon fall back
 # to the plain Common of a Confessor. No Lectio7-9 in his own file --
 # defers to the Common's own Nocturn 3, not transcribed here.
+#
+# Deliberately demoted from the reference engine's own Duplex to
+# Semiduplex (2026-08 user decision) -- a minor confessor of no
+# particular standing shouldn't outrank the Assumption's own octave,
+# now strengthened to `wins: duplex` (see assumption.yml) specifically
+# so this demotion actually takes effect on his day (Aug 19, the
+# octave's 5th day) instead of merely tying and still winning.
 id: st-john-eudes
 name: "St. John Eudes, Confessor"
-rank: duplex
+rank: semiduplex
 common: common-of-a-confessor
 propers: "st-john-eudes"
 benedictusCommon: common-of-a-confessor

src/data/calendar/saints/st-thomas-becket.yml

@@ -6,9 +6,15 @@
 # `privileged-feria` model would have wrongly forbidden). Octave day
 # commemorated in return. Own proper collect, no unique antiphon of his
 # own.
+#
+# Deliberately promoted from the reference engine's own Semiduplex to
+# Duplex (2026-08 user decision, made alongside assumption.yml's octave
+# strengthening to `wins: duplex`) -- purely to keep him winning Dec 29
+# outright once the common octave-vs-saint default is no longer
+# uniformly Semiduplex everywhere; not a reference-engine finding.
 id: st-thomas-becket
 name: "St. Thomas of Canterbury, Bishop and Martyr"
-rank: semiduplex
+rank: duplex
 common: common-of-a-martyr-bishop
 propers: "st-thomas-becket"
 # P/T/S/N: no [Ant Prima]/etc of its own in the source (12-29o.txt) --