Complete the aging chain for rebased capital-gains state rows (attempt-3 adjudication)

[MaxGhenis]

Adjudicates Build N attempt 3's only national gate failure. Full diagnostic receipts on #462 (raw facts, factor sources, and closure identities, all with feed/workbook line references).

The defect: the ht2 TY2022 state net-capital-gains family rebases onto the TY2023 Table 1.4 control (×0.771900, arithmetically exact for its inputs) and then stops — the rebase stamped uprating_to_period=2024 while landing at TY2023 data, and the ager skipped any row carrying uprating_factor as "already aligned". The national row continued TY2023→TY2024 via CBO (×1.315366, source-exact against the CBO workbook). Fifty-one mutually consistent state rows at TY2023 levels faced one national row at TY2024, 34% apart; the solve satisfied the 51 and ate −25.58% on the national — byte-identical to certified Build M's known gap, now blocking under #464. An input rescale (the intuitive fix) would have broken 51 excellent rows to satisfy one.

The fix keeps every chain link exactly once:

• the CG rebase stamps uprating_to_period with the control's actual period;
• target aging treats uprating_to_period as an uprated row's effective source period — rows uprated to build fall through to source_equals_build at factor 1 (values bit-identical; the already_period_aligned reason string retires), rows stopped at an earlier control receive exactly the remaining CBO links.

Closure identities (from the adjudication): 51-state sum → ~$1,245.1B vs national $1,270.9B; the $25.75B residual matches the exactly-identified OA+PR slice ($25.37B = ht2 US minus the 51 states, proven row-by-row from the source CSV) — internal consistency to ~0.03%. Also recorded there: the ht2-combined vs Table-1.4-gains-only concept distinction (0.17% same-year effect; ledger concept-alignment follow-up) and the fact that the other three uprating writers stamp the build period by construction (byte-identical behavior here; stale-control audit for them is a follow-up).

Tests: chain-completion law pinned (a TY2023-control row receives exactly the 2023→2024 CBO net-capital-gain factor, aged from the control, nothing twice); the rebase's control-period stamp pinned; the double-age guard updated to the truthful reason. Full test_us_fiscal_targets.py plus builder/plan/parity/SSI suites green locally.

Expected on attempt 4: the state family binds ~31.5% higher, the national row becomes honestly reachable, and the solve reconciles the $25B coverage sliver across both families (~1% each, inside all tolerances). Part of #462/#464.

🤖 Generated with Claude Code

[MaxGhenis]

Round-2 disposition: finding 1 fixed in 5cf06ce — the malformed-stamp guard now runs for every row carrying uprating_factor, ahead of the not-ageable early return, with a count-row regression (counts never age, but a populace-authored stamp that doesn't parse fails closed). Finding 3 verified passing in round 2 (model 1.2.0 on all 2,007 aged specs, zero 1.1.0 pins).

Finding 2 is adjudicated out of this PR as pre-existing and improved by it: the repaired AGI-bin sub-surface closes to its own CD-lineage control at 0.020%; the residual 9.76% gap is between the HT2 national all-row and the bins' partition — a source-level HT2-vs-CD-table discrepancy that existed on main (~20% there, since the bins also sat one CBO link behind) and that no blocking gate binds. Filed with the round-2 receipts as #489 alongside the EITC group-stamp audit.

Verified value delta remains exactly the adjudicated classes: 51 CG rows ×1.3153658 + 467 interest rows ×1.0872135, 518 changed / 5,697 total, nothing else moved. Merging on green CI; attempt 4 runs against the vintage-consistent surface.