CG depth charter: close the dense netCG gap (−30.8%, B) so the dense arm can publish

Chartered by Max 2026-07-28 ("charter cg depth yes"). The dense arm's sole structural fit blocker: table_1_4.all.net_capital_gains_amount target $1,270.9B, dense final $879.1B — −30.8%, invariant across two bases and three attempts (byte-level agreement between Build P2 attempts 2/3; dense-P at base-P: −30.99%). The sparse arm passes the same row via rmloss100 L0 concentration; the ratio-bounded full-pool solve cannot reach it. Receipts: #507 attempts 2/3.

Phase 1 — diagnostic (sol lane RUNNING)

1. Decompose the shortfall on the dense frame at final weights: carrier composition, participation vs magnitude vs top-tail.
2. Aggregate-record hypothesis: the raw TY2015 PUF's four MARS==0 aggregate records carry the extreme tail and the pipeline drops them (proven for NOL: they absorb most of SOI's $197.5B). Measure raw CG fields regular-vs-aggregate, weighted, vs SOI 1.4 ty2015; then donor and frame. If the pool structurally lacks the top CG tail, quantify the reachable ceiling under the current ratio bound.
3. How concentrated is the sparse arm's passing CG mass (top-100 share)? Bears on pool-defect vs arm-design.

Phase 2 — design (on diagnostic receipts)

Candidates to cost, none pre-committed: truthful aggregate-tail representation at base-build (provenanced synthetic tail records — inverse-#462 spirit: adding measured mass, not clamping); CG-specific QRF tail-depth work (donor-vs-frame tail quantiles); SCF blend leverage (scf_wealth already ON).

Anti-charter (standing doctrine): no loss shaping, no per-target multipliers, no gate/envelope changes, no max_weight_ratio bumps as "the fix".

Exit criterion

A dense run whose netCG row passes the existing 0.25 gate on truthful inputs — no gate moved. Unblocks dense-P3 publication alongside #566 (dense take-up threshold design).

Related: #487 (cancelling-legs class), #462 (donor-tail machinery), #566, #507 receipts.

Start with the diagnostic receipts from #507 and the dense-frame final weights. Compare raw TY2015 PUF regular and aggregate records against SOI 1.4 TY2015, then measure donor and frame tail coverage and the sparse arm's top-100 share. Done means a dense run passes the existing 0.25 netCG gate using truthful inputs, without changing the gate or listed anti-charter controls.