Function render_active 

pub fn render_active(
    rules: &[&Rule],
    tallies: &BTreeMap<String, RuleTally>,
) -> String

One line per active rule, carrying what the validation ledger measured about it — the learner’s evidence for what to drop.

attributed is the number retirement argues from: a bisection pinned that regression on this rule specifically. Block-level improved/regressed only say what the whole rule set did while this rule rode along, which is context rather than credit.

Unmeasured is rendered as unmeasured, never as zero. A rule no probe has covered has no evidence either way, and printing 0 regressed would read as a clean bill of health while printing nothing would read as harmless. Both are the “a dash is never zero” failure, and collapsing them would bias the learner against the newest rules — the ones with least chance to have been probed.