test(oracle): fix round 1 -- C30 disclosure, RG 17(a) misattribution, M26 shapes

Three Important findings from coordinator review of 8d1eb25, all fixed.

1. C30's own note claimed its 91-row population "is C25's OWN former
   population, before this task narrowed it" -- false. Independently
   re-derived against the real resolver (pre-fix vs post-fix binary) and
   the lectio fixture: only 44 of the 91 rows were ever in C25 (Wed 6/
   Thu 19/Fri 19, matching C25's own fix-round-3 breakdown exactly); the
   other 47 (all of Monday/Tuesday, plus part of Wednesday/Friday)
   previously matched lectio outright -- both engines independently
   landing on the same wrong value -- and are newly allow-listed only
   because colitur is now right. Rewritten to the same "both engines
   silently agreed on the wrong answer" pattern C33/C34 already used.

2. "Diebus ferialibus" is not RG 17(a) -- RG 17(a) states only the
   day-identity privilege; the ferial-Mass rubrics are Mass-propers
   rubrics, a distinction the register's own §6.3 already drew before
   this task's first pass blurred it in code/allow-list comments. A
   further slip: C31/C32 called their own rubric "RG 17(a)'s own next/
   third sentence" when it is a different rubric block ~230 lines away,
   in Holy Family's (not Holy Name's) own feast propers. Fixed the
   attribution chain throughout tools/bootstrap_lectionary.ml,
   data/ef/expected-divergences.sexp's C29-C34, and (gitignored, outside
   this commit) docs/research/rules-register.md -- closed the §6.3 park,
   added §6.5 recording all four citations with corrected attribution.
   Quotes and line numbers were already right; only the chain of custody
   to RG 17(a) was wrong.

3. M26's own 30 rows classify into two nameable shapes, not "three
   confirmed shapes" with an unexplained 23-row residue that was in fact
   empty: 21 rows are RG 78's own already-documented BVM votive-Mass
   gap; 9 rows are a second, distinct gap -- a week whose own Sunday was
   displaced by a movable named feast (Trinity->Corpus Christi, "Missa
   Dominicae I post Pentecosten"; Christ the King), where the Missal
   requires the displaced Sunday's own Mass to keep informing that
   week's ferias but step 3 walks back to the displacing feast's citation
   instead -- proven directly by colitur's own citation on the Christ-
   the-King rows being literally that feast's own Gospel. verdict open
   unchanged (not fixed here); the shape is now named instead of buried.

Minors: section count corrected 1458->1456 (four places); "exactly one
candidate, zero ambiguous" corrected to "correctness rests on ordering,
not uniqueness" (2 of 1456 sections genuinely yield two candidates,
Holy Saturday's own Lectio, both years); the stale "nothing in layers
3-5 compares citations" claim corrected (only layer 5 still doesn't);
M18's note corrected (five new entries not six; ~18 of its remaining
352 rows only agree now because of this task's own fixes, not "already"
before it); "five" corrected to "six" otherwise-unoccupied Saturdays in
bootstrap_lectionary.ml; scan2 quotes softened where OCR-garbled.

Also done outside the git-tracked files: the step-4 (Commons) coverage
gap is now recorded in data/ef/expected-divergences-missalemeum.sexp's
own header, not only the commit body and the (gitignored) report.

dune test: 358 tests green. COLITUR_EXHAUSTIVE_SWEEP=1 dune test
--force: 359 tests green, 95.7s.

3 files changed, 112 insertions, 63 deletions

diff --git a/test/fixtures/missalemeum-ef-2026-2027.provenance b/test/fixtures/missalemeum-ef-2026-2027.provenance
index d760c9f..33b5566 100644
--- a/test/fixtures/missalemeum-ef-2026-2027.provenance
+++ b/test/fixtures/missalemeum-ef-2026-2027.provenance
@@ -49,11 +49,24 @@ things verified during extraction, not merely assumed: no field collides
 with the "|"/";" delimiters, all 4 multi-Mass days (info is an array on
 2026-11-02, 2026-12-25, 2027-11-02, 2027-12-25) carry an IDENTICAL info
 block (rank/colors/tempora/commemorations/displaced) across every Mass of
-the day, so taking entry[0] throughout (this fixture's own convention)
-loses nothing the comparator or a human auditor needs, and the citation
-extraction itself (all 1 458 Lectio/Evangelium sections across the 728
-days that carry one) yields exactly one unambiguous citation-shaped
-candidate per section -- checked exhaustively, not sampled.
+the day, so taking entry[0] throughout for THOSE fields (this fixture's
+own convention) loses nothing the comparator or a human auditor needs
+there -- CORRECTED, fix round 1 (coordinator review): this is NOT true
+for the citation fields Task 9 added. Christmas (25 December) has three
+genuinely different Masses (Midnight/Dawn/Day) with three different
+Epistle+Gospel pairs; entry[0] is the Midnight Mass, colitur's own
+citation (bootstrapped from lectio) reflects the Day Mass -- a real,
+documented divergence (test_oracle.ml's own M27), not one this
+convention avoids. The citation extraction itself (all 1 456 Lectio/
+Evangelium sections, 728 days x 2) does NOT yield exactly one candidate
+per section either -- CORRECTED, fix round 1: 2 of the 1 456 (Holy
+Saturday's own "Lectio", both years) yield two verse-shaped spans,
+['Col 3:1-4', 'Ps 117, 1'] (the second is the Tractus verse following
+the real citation in the same paragraph). Both rows are still extracted
+correctly because [extract_citation] returns candidates[0] and the real
+citation happens to come first -- correctness rests on ORDERING, not
+uniqueness. See tools/extract_missalemeum_oracle.py's own header for
+the full account of both corrections.
 
 Regenerate with the same commands if the missalemeum snapshot is refreshed
 and the fixture needs updating; do not hand-edit this file or the fixture.
diff --git a/test/test_golden.ml b/test/test_golden.ml
index 2f8930f..8d33e29 100644
--- a/test/test_golden.ml
+++ b/test/test_golden.ml
@@ -94,11 +94,15 @@ let real_ef_lectionary =
 
 (* The Commons (data/ef/commons.sexp) travel the same caller-supplied seam
    as the lectionary above, and [~commons] is required rather than defaulted
-   so that no caller can silently run with none -- nothing in layers 3-5
-   compares reading citations, so a rite quietly missing its Commons would
-   be invisible. Loaded here even where this file asserts nothing about
-   readings, so that the rite under test is the same one bin/main.ml
-   assembles. *)
+   so that no caller can silently run with none. CORRECTED, Task 9 fix
+   round 1 (coordinator review): this used to claim "nothing in layers 3-5
+   compares reading citations" -- stale since Task 8 (layer 3) and Task 9
+   (layer 4). THIS layer (5, golden pins) is the one that still does not:
+   this file's own [describe] carries no citation field at all, so a rite
+   quietly missing its Commons would still be invisible here specifically,
+   even though layers 3/4 would now catch it. Loaded here even where this
+   file asserts nothing about readings, so that the rite under test is the
+   same one bin/main.ml assembles. *)
 let real_ef_commons =
   match Rite_ef.Lectionary_ef.Commons.load commons_path with
   | Error e -> failwith (Printf.sprintf "%s: failed to load: %s" commons_path e)
diff --git a/test/test_oracle.ml b/test/test_oracle.ml
index 41d8a2e..1ad4b1f 100644
--- a/test/test_oracle.ml
+++ b/test/test_oracle.ml
@@ -234,9 +234,13 @@ let real_lectionary () =
 
 (* The Commons (data/ef/commons.sexp) travel the same caller-supplied seam
    as the lectionary above, and [~commons] is required rather than defaulted
-   so that no caller can silently run with none -- nothing in layers 3-5
-   compares reading citations, so a rite quietly missing its Commons would
-   be invisible. Loaded here even where this file asserts nothing about
+   so that no caller can silently run with none. CORRECTED, fix round 1
+   (coordinator review): this used to claim "nothing in layers 3-5 compares
+   reading citations" -- stale since Task 8 (layer 3) and this task itself
+   (layer 4, this file). Layer 5 (test_golden.ml's own [describe]) still
+   has no citation field at all, so it alone would still miss a rite
+   quietly missing its Commons; layers 3/4 compare citations directly and
+   would not. Loaded here even where this file asserts nothing about
    readings, so that the rite under test is the same one bin/main.ml
    assembles. *)
 let real_commons () =
@@ -1161,57 +1165,85 @@ let m24_dates = [ "2026-04-01"; "2026-04-02"; "2027-03-24"; "2027-03-25" ]
    predicate already applies to a citation-vocabulary gap. *)
 let m25_dates = [ "2026-03-09"; "2026-03-24"; "2027-03-01"; "2027-03-16" ]
 
-(* M26 -- Task 9: the large remaining population, one single root cause.
-   {!Rite_ef.Lectionary_ef.readings}'s own header comment already names
-   this LIMIT, not a rubric dispute this task settled: step 3 ("a feria
-   with no proper says the preceding SUNDAY's Mass") is colitur's only
-   fallback mechanism, and its own WARRANT is "lectio's own observed
-   behaviour... not a confirmed Missal citation" -- register's own
-   standing note. missalemeum's own ferial-Mass selection is more
-   granular than that single rule in (at least) three distinct ways this
-   task's own investigation found and confirmed against the primary
-   source, but did NOT build (a genuine, understood, ALREADY-DOCUMENTED
-   architecture gap, the SAME shape data/ef/expected-divergences.sexp's
-   own C6/C25 entries already carry for the sibling lectio differential,
-   and Task 6's own residual concern 4/5, "the BVM Saturday Office still
-   emits its feria's Mass... needs... a season-keyed selection among five
-   formularies" -- RG 309(a), "iuxta temporum diversitatem"):
-   (1) a rotating VOTIVE MASS OF THE BVM ("Missae de sanctae Mariae in
-   sabbato" I-V and their Time-after-Pentecost siblings, missalemeum's own
-   "II/III/V Mass of the B. V. M." titles) on an otherwise-unoccupied
-   feria, not only a Saturday -- 2026-01-03/2026-01-10/2027-01-02/
-   2027-01-09 are RG 78's own BVM SATURDAY Office specifically (Task 6's
-   own named open item; colitur's own step-3 fallback, not the votive
-   Mass, is what these four rows actually show -- 2026-01-03/2027-01-09
-   are `ef-christmas-1-saturday`, C25's own unchanged population in the
-   sibling lectio differential; 2026-01-10/2027-01-02 are
-   `ef-christmas-2-saturday`); the rest of this entry's own population
+(* M26 -- Task 9: the remaining population, one root cause,
+   TWO NAMEABLE SHAPES -- CORRECTED, fix round 1 (coordinator review,
+   Important 3): a prior version of this entry described three shapes,
+   the third an unexplained "23-row residue... not narrowed to a single
+   citable rule at all". That bucket was EMPTY -- every one of the 30
+   dates classifies cleanly into one of the two shapes below; the
+   "residue" framing buried a real, nameable defect instead of recording
+   it. {!Rite_ef.Lectionary_ef.readings}'s own header comment already
+   names the underlying LIMIT, not a rubric dispute this task settled:
+   step 3 ("a feria with no proper says the preceding SUNDAY's Mass") is
+   colitur's only fallback mechanism, and its own WARRANT is "lectio's
+   own observed behaviour... not a confirmed Missal citation" --
+   register's own standing note. missalemeum's own ferial-Mass selection
+   is more granular than that single rule in two distinct, CONFIRMED ways
+   (the same "needs a season-keyed selection mechanism, new behaviour not
+   new data" shape data/ef/expected-divergences.sexp's own C6/C25 entries
+   already carry for the sibling lectio differential, and Task 6's own
+   residual concern 4/5, "the BVM Saturday Office still emits its feria's
+   Mass... needs... a season-keyed selection among five formularies" --
+   RG 309(a), "iuxta temporum diversitatem"):
+
+   SHAPE 1 -- 21 rows: a rotating VOTIVE MASS OF THE BVM ("Missae de
+   sanctae Mariae in sabbato" I-V and their Time-after-Pentecost
+   siblings, missalemeum's own "II/III/V Mass of the B. V. M." titles) on
+   an otherwise-unoccupied feria, not only a Saturday -- 2026-01-03/
+   2026-01-10/2027-01-02/2027-01-09 are RG 78's own BVM SATURDAY Office
+   specifically (Task 6's own named open item; colitur's own step-3
+   fallback, not the votive Mass, is what these four rows actually show
+   -- 2026-01-03/2027-01-09 are `ef-christmas-1-saturday`, C25's own
+   unchanged population in the sibling lectio differential; 2026-01-10/
+   2027-01-02 are `ef-christmas-2-saturday`); the other 17 of these 21
    are ordinary weekday ferias in Time after Pentecost showing the SAME
-   votive-Mass title;
-   (2) a NAMED weekly formula distinct from the plain Sunday-repeat rule
-   for at least one week this task's own primary-source reading found by
-   accident while fixing a DIFFERENT gap (Holy Family's own week,
-   "Missa Dominicae Primae post Epiphaniam" -- built, tools/
-   bootstrap_lectionary.ml's own [epiphanytide_opening_entries]) --
-   2026-06-02/03 and 2027-05-24 are the SAME shape for a different week
-   (the ferias between Trinity Sunday and Corpus Christi, "Missa
-   Dominicae I post Pentecosten", 1 John 4:8-21/Luke 6:36-42, visible in
-   the raw scan while researching Corpus Christi's own Mass -- see
-   docs/research/scan1.txt around line 21800 -- but NOT independently
-   verified as a general "every week has its own named formula" rule the
-   way the January window was, and NOT built here);
-   (3) later Time-after-Pentecost weeks whose own votive/ferial answer
-   this task's investigation did not narrow to a single citable rule at
-   all (2026-10-26 through 2026-11-28, 2027-11-03 through 2027-11-27).
-   ALL THREE are variations on the identical root limitation, not three
-   separate bugs, and none is safe to fix piecemeal inside this task's own
-   scope (extending layer 4 + triage) without the same real, dedicated,
-   TDD/mutation-tested unit of work C6/C25 and Task 6's own open item
-   already call for -- a season-keyed ferial/votive-Mass-selection
-   mechanism, new behaviour, not new data. Verdict open, the SAME fourth
-   verdict M15/M18 already establish for a genuine comparator/architecture
-   LIMIT, honestly counted rather than silently absorbed or left failing
-   without explanation. A literal date list, not a shape predicate (no
+   votive-Mass title (2026-02-14/06-20/06-27/07-04/07-11/08-01/10-31/
+   11-07/11-28; 2027-07-17/07-24/09-04/09-11/10-30/11-06/11-13/11-27).
+
+   SHAPE 2 -- 9 rows, a SECOND, DIFFERENT and DISTINCT gap: ferias of a
+   week whose own Sunday was DISPLACED that year by a movable, named
+   temporal-cycle feast, where the Missal requires the DISPLACED
+   Sunday's own Mass to continue informing that week's ferias, but step 3
+   walks back to the DISPLACING feast's citation instead (the SAME
+   Sunday's own OFFICE is displaced; its own reading is not). Two
+   witnessed instances, not a residue: (a) 2026-06-02/03 and 2027-05-24,
+   the ferias between Trinity Sunday and Corpus Christi -- the Missal
+   prints a named formula for exactly this week, "Missa Dominicae I post
+   Pentecosten" (1 Ioann. 4,8-21 / Luc. 6,36-42, docs/research/
+   scan1.txt:21758-21830, found while researching Corpus Christi's own
+   Mass immediately following it) -- but colitur's step 3 walks back to
+   Trinity SUNDAY's own citation (Rom 11:33-36, the generic Time-after-
+   Pentecost fallback), not this named formula, since [ef-corpus-christi]
+   now correctly interrupts Trinity's own week at step 2 (this task's own
+   fix) and step 3's walkback from the Tuesday/Wednesday after it lands
+   one Sunday further back than the week actually calls for; (b)
+   2026-10-26/27/29/30 and 2027-11-03/05, the ferias of the week Christ
+   the King (I class, movable, "Dominica ultima octobris") displaces --
+   colitur's own citation on all six is LITERALLY Christ the King's own
+   Gospel (John 18:33-37, "Art thou a king?"), confirming step 3 walked
+   back to the DISPLACING feast, not the displaced ordinary Sunday;
+   missalemeum shows a different citation each row (the displaced
+   Sunday's own, varying by year). No Missal-propers rubric for Christ
+   the King's own week ferias was found the way Holy Name/Holy Family's
+   own each carry one (checked: no rubric text precedes Christ the
+   King's own Mass heading, scan1.txt:39140-39163, unlike the two
+   feasts C30/C31/C32 above document) -- this half is named from the
+   OBSERVED citation evidence itself (colitur's own value IS the
+   displacing feast's, missalemeum's is not), not from a located
+   propers rubric; the general mechanism (a week's own ferias follow its
+   OWN Sunday's temporal identity even when that Sunday is impeded that
+   year) is the same shape RG 69 already establishes for the Nativity
+   Octave (C6's own citation) and RG 96's own general transfer framework
+   assumes throughout, but no citation SPECIFIC to Christ the King's own
+   week was found or is claimed here.
+
+   Neither shape is safe to fix inside this task's own scope (extending
+   layer 4 + triage) without the same real, dedicated, TDD/mutation-tested
+   unit of work C6/C25 and Task 6's own open item already call for.
+   Verdict open, the SAME fourth verdict M15/M18 already establish for a
+   genuine comparator/architecture LIMIT, honestly counted and NAMED
+   rather than silently absorbed, left failing, or buried in an
+   unexplained residue. A literal date list, not a shape predicate (no
    single structural signal -- unlike C17's own [subject = bvm] -- exists
    to distinguish this population from a genuinely new, unrelated
    citation bug on some other Time-after-Pentecost feria): the SAME