Historical compute split: The active-49 planning record preceded the final 47-primary plus 2-coverage publication and remains provenance only.
Trial 62 keeps the original 84-row, 14-cluster research family and all 66 canonical implementations unchanged. Batch 1 executes exactly 49 rows: 48 SFP-compatible strategies plus W9-103. The 17 implemented broad-SEP rows W9-001, W9-002, W9-003, W9-004, W9-009, W9-025, W9-033, W9-034, W9-035, W9-056, W9-083, W9-087, W9-092, W9-094, W9-107, W9-113, and W9-114 are deferred by owner decision for compute cost before any broad-row result. The remaining 18 rows retain their existing capability exclusions.
The trial-62 amendment supersedes the historical 40 Stage-1 / 38 Stage-2 / 6 data-held staging, its PROVISIONAL label, and its wait-for-Stage-2 final-correction rule. Those dated records remain provenance only. “ONE batch” describes the preserved original84 historical search family and multiplicity burden; active49 is the complete independently publishable batch-1 correction set and does not wait for a deferred batch-2 result.
The sealed reconciliation is therefore 49 active + 17 deferred + 18 excluded = 84. All 66 source modules and registered hashes are still loaded and verified, and unknown modules still refuse admission. Batch-1 PBO and Holm/BH use all 49 evaluated rows across 8 nonempty execution clusters. Historical DSR remains charged at raw N=84 and 14 clusters; the ledger continues to declare batchN=84, familyClusters=14, and promotableVariants=0.
Publication, correction, combination, and candidate selection require the exact active49 roster. Missing, extra, deferred, excluded, or trial-61 rows refuse. Deferred rows remain implemented and accounted for pending a separately sealed batch-2 registration. No survivor receives automatic paper or live authority: it still requires fresh single-variant preregistration that discloses this exploration.