# Reproduced Second Close results

All values are exact rational modeled USD. Case inputs are constructed.

26 cases; 1024 complete policy trees independently enumerated.

| Case | Optimal value | Every maximizing first action | Trees |
|---|---:|---|---:|
| [authority-withdrawn](../cases/authority-withdrawn.json) | -8 | reconcile | 4 |
| [cheaper-evidence](../cases/cheaper-evidence.json) | -2 | investigate | 41 |
| [cheaper-production](../cases/cheaper-production.json) | -2 | investigate, produce, reconcile | 41 |
| [congestion-blocked](../cases/congestion-blocked.json) | 0 | decline, defer | 2 |
| [congestion-cleared](../cases/congestion-cleared.json) | 4 | produce | 4 |
| [conventional-equal-facts](../cases/conventional-equal-facts.json) | -3 | investigate | 41 |
| [correlated-effects](../cases/correlated-effects.json) | -6 | reconcile | 2 |
| [costly-investigation](../cases/costly-investigation.json) | -4 | produce, reconcile | 41 |
| [costly-resolution](../cases/costly-resolution.json) | -6 | defer, produce | 41 |
| [deferred-loss-omitted-invalid-control](../cases/deferred-loss-omitted-invalid-control.json) | 4 | produce | 4 |
| [deferred-production-loss](../cases/deferred-production-loss.json) | 0 | decline, defer | 4 |
| [horizon-1](../cases/horizon-1.json) | -6 | produce | 5 |
| [horizon-3](../cases/horizon-3.json) | -1 | investigate, produce | 540 |
| [independent-effects](../cases/independent-effects.json) | -10 | defer | 2 |
| [joint-loss-correlated](../cases/joint-loss-correlated.json) | -6 | reconcile | 2 |
| [joint-loss-independent](../cases/joint-loss-independent.json) | -5 | defer | 2 |
| [known-bad](../cases/known-bad.json) | -8 | reconcile | 5 |
| [known-good](../cases/known-good.json) | 4 | produce | 5 |
| [qao-capacity-2](../cases/qao-capacity-2.json) | -6 | defer, produce | 25 |
| [qao-capacity-3](../cases/qao-capacity-3.json) | -3 | investigate | 41 |
| [repartitioned](../cases/repartitioned.json) | -3 | investigate | 41 |
| [settlement-before-evidence](../cases/settlement-before-evidence.json) | -6 | produce | 4 |
| [slack-capacity](../cases/slack-capacity.json) | -3 | investigate | 41 |
| [successor](../cases/successor.json) | -3 | investigate | 41 |
| [wait-for-settlement](../cases/wait-for-settlement.json) | 0 | defer | 5 |
| [withdrawal-after-wait](../cases/withdrawal-after-wait.json) | -3 | investigate | 40 |

## Capacity investment

| Capacity | Gross gain | Upfront cost | Net gain | Decision |
|---|---:|---:|---:|---|
| 2 → 3 | 3 | 2 | 1 | invest |
| 2 → 3 | 3 | 3 | 0 | indifferent |
| 2 → 3 | 3 | 4 | -1 | retain |
| 3 → 99 | 0 | 2 | -2 | retain |

The positive case is a constructed value of physical resolution capacity. It is not measured Bulla ROI or provider revenue.

Production-summary witness: free-signal value -2; coarse-summary value -6; expected regret 4.

See results.json for every root action value, contingent policy and source digest. The deliberately misspecified continuation case demonstrates that exact reconciliation does not validate a premise.