Hybrid Rare-Event Estimation for Conjunction Risk: Benchmark Results and Fault-Tolerant Resource Sketches

Satellite conjunction assessment relies on repeated probability-of-collision (Pc) estimation under uncertainty as new tracking updates arrive. For most conjunctions, coarse screening is sufficient, but decisions near maneuver thresholds require rapid time-to-confidence in tail probability, where brute-force Monte Carlo becomes expensive and sensitive to distributional assumptions. This report presents a reproducible encounter-plane benchmark suite that stress-tests rare-event estimators across baseline and near-threshold regimes, including mixture (non-Gaussian) uncertainty modes. We compare naive Monte Carlo, importance sampling, and a splitting-style rare-event estimator using confidence-interval width versus sample budget as the primary metric. We outline how the same event-predicate structure supports quantum amplitude-estimation-style acceleration in a hybrid workflow, and we provide initial fixed-point precision and fault-tolerant resource sketches. Code and scripts to reproduce all figures are available in the linked repository.