Audit-Ready Path Promotion for CL-QUBO Builder Families: A Benchmark, Comparison, and Freeze Decision for Fault Tree Minimal Cut Set Encodings

This paper presents an audit-ready benchmark, comparison, and freeze decision for alternative CL-QUBO builder and variational family paths applied to fault tree minimal cut set (MCS) identification. The study evaluates whether cost-phase-augmented variational families from prior work can challenge the incumbent CL-QUBO builder path (v5) on a widened 9-case tiny validation corpus covering OR, AND, NOT, and nested gate structures. A disciplined promotion pipeline, comprising role mapping, contract extraction, execution coverage adjudication, semantic grid screening across 1, 305 parameter configurations, and formal synthesis, was exercised to ensure that challengers were genuinely advanced rather than dismissed prematurely. Two evidence tiers structure the comparison: the incumbent v5 was evaluated through exact combinatorial audit (exhaustive projected-MCS-family recovery), while challengers were evaluated through best-of-grid statevector screening (top-1 MCS family membership and total MCS mass). The incumbent v5 achieved 9 of 9 primary target hits (minimum-cardinality MCS under the minimize-basic-events objective) and 8 of 9 exact all-MCS family recovery hits, with the single miss occurring on a case where the minimum-cardinality MCS is a strict subset of the full MCS family. The independent cost-phase challenger achieved 3 of 9 primary hits and 3 of 9 all-MCS top-1 hits at best-of-grid. The block-shared cost-phase challenger achieved 3 of 8 primary and 3 of 8 all-MCS top-1 hits on applicable cases, with one case (TVC₀07NOTA) structurally not applicable because the block-shared generator schedule requires a retained prefix structure that the one-qubit case does not provide. The completion gate passed all eight checks, and v5 was frozen as the final promoted path within the widened tiny benchmark scope studied. This paper does not claim quantum advantage.