Bootstrap each conditional metric over its defined replicas

A conditional metric may be undefined in bootstrap replicas where its eligible denominator is empty. Do not index every metric by the total requested replica count or silently coerce undefined values to zero. Accumulate defined values independently for each metric, compute its interval from that metric’s actual replicas, and report the defined-replica count beside the interval. This preserves the estimand while making sparse support visible. If too few replicas define the metric, withhold the interval rather than implying precision.

Applies when: bootstrapping route-conditioned, subgroup, precision, recall, or gated evaluation metrics. Keywords: bootstrap conditional metric empty denominator confidence interval sparse subgroup replicas