# C332 independent final selection diagnostic

24 September 2026. The audit retains exactly C319's20 controls: i3…7 and independent cumulative/regular choices a,b∈{0,2}. The separately admitted C287 i2 row is not added to this menu. N, its matched SP death D and actual same-Gear restored-MT Noah M are independently retrieved from C286. Cumulative S and regular R membership are rechecked rather than assumed from their formal coordinates.

**Result:** the death partition does not independently force the paired22/all-K choice. `950+108` forces b=2 but leaves a free; requiring490 also restricts the SP primary rail. Two source-admitted02 controls satisfy both new conditions while failing the all-K form. Adding the existing two17K condition leaves22 primaries i3/5, but that uniqueness is conditional on the pre-existing coupling requirement.

## Exact conditions and source-qualified counts

The complete finite formulas are:

`N−D=950`, `D−R=110−b`, and `M−D=490` on primary rails or491 on companion rails.

Thus950+108 is exactly the already held right-side2K condition, because `N−R=1060−b`. Neither new regular interval involves S or the cumulative choice a. The actual MT endpoint remains an ordinary admitted birth; no companion is invented to make491 become490.

| Criterion | Formal controls /20 | Jointly source-admitted controls /14 |
|---|---:|---:|
|950+108|10|6|
|490 alone|12|9|
|Both new conditions|6|4|
|Both new conditions plus existing two17K|3|2|
|Existing all-K form|5|3|

These are counts of control configurations, which can reuse the same regular endpoints; they are not independent witness counts.

| a /b | Jointly admitted i |950+108 admitted i|490 admitted i|Both admitted i|Both plus two17K admitted i|
|---|---|---|---|---|---|
|0 /0|3,4,5,6,7|—|3,5,7|—|—|
|0 /2|3,4,5|3,4,5|3,5|3,5|—|
|2 /0|3,4,5|—|3,5|—|—|
|2 /2|3,4,5|3,4,5|3,5|3,5|3,5|

The source-valid counterexamples to independent22 selection are `(a,b,i)=(0,2,3)` and `(0,2,5)`. At i3 the existing chain is3706→3246→2756→2648, with460+490+108, while S=−6347 makes S→R8995 rather than8993. The analogous i5 control also passes both new regular tests. The differing cumulative S corner is not constrained by those regular intervals.

The stronger conjunction must also be stated positively: existing two17K requires a=b; the108 condition requires b=2; together they require a=b=2. Adding490 leaves the two admitted primary rows i3/5. This is a correct conditional intersection. It restates the existing right2K and paired-choice requirements, rather than establishing an independent source instruction to select them.

Conversely, the retained all-K companion `(2,2,4)` has491, not490. A490 requirement excludes a source-valid all-K configuration; it is not a necessary property of every all-K row.

## Admission and phases

R is checked as MT Gear `g+b/2`, start for a primary SP rail and close for a companion rail, in restoredON/Terah60/full430. S is checked as cumulative ON index `i+a` paired with held OFF Flood_i. At the upper controls, required Gear4 Flood endpoints and/or cumulative members8/9 remain absent. Their six formal-only configurations stay in the data with explicit failed membership; formal arithmetic is not promoted to source admission.

Only C319's two held phase components per row are reused. M and D receive that same physical output offset. All40 phase configurations preserve their annual diagnostic values;28 are jointly source-admitted. No wider phase menu, ancestry identification or missing-source repair is introduced.

## Reproduction

`C332_Independent_Audit.py --sources DIR --out DIR` uses Python's standard library and frozen C319 data/report plus C286. Its JSON retains all20 controls,40 components, source records and failures, exact predicate counts, explicit counterexamples and the stronger two17K survivors. **432 checks passed.** The script, results and review are frozen for handoff. C333 remains synthesis only; no additional research has been executed here.