Overturn-rate reputation mixes round-one and appeal populations

applySlashAndReward runs unconditionally for both round zero and appeal rounds, incrementing casesResolved for every revealed voter, while appealOverturnCount can only ever be incremented for round-one jurors. The eligibility ratio therefore uses a denominator inflated by appeal-only participation: a juror excluded at a 2/10 overturn rate becomes eligible again at 2/13 after three appeal-only reveals, with no improvement in their round-one accuracy.

The juror must first obtain appeal selections and may already be excluded, so the dilution is neither immediate nor permissionless.

Track round-one resolved votes and round-one overturns in one consistent denominator, and maintain separate appeal metrics if appeal performance is to be measured.

Sequence this after the tally-layer correction: fixing the numerator, which currently counts false overturns, matters more than fixing the denominator, and that change alters what this metric measures.

Fixed. casesResolved increments only on round-zero tallies, so appeal-only participation no longer dilutes the overturn denominator.

• lib/ArbitrationPoolTallyLib.sol#L169-L174, #L239-L255
• lib/ArbitrationPoolSelectLib.sol#L171-L173