A validation report shows the transformed Java rounds a currency field differently than the legacy COBOL did. What does this finding mean for the team?

A rounding difference on a currency field is precisely the kind of subtle divergence comparative testing exists to catch. The finding means the transformation introduced a real behavioural difference, and the generated code must be corrected so its rounding matches the legacy program's intended logic. The suggestion that the finding can be closed once the new rounding behaviour is documented as the accepted behaviour going forward is incorrect because documenting the new behaviour as acceptable abandons the requirement to preserve the original business logic rather than satisfying it. The view that the finding is informational only, since rounding differences rarely affect downstream calculations, is incorrect because currency rounding errors compound across transactions and reporting periods, and are rarely harmless. The recommendation to resolve the finding by loosening the test suite's numeric tolerance until it stops flagging the case is incorrect because loosening tolerance to silence the failure hides the defect instead of resolving it. For the exam, financial calculations are exactly where small unexplained differences need investigating, not tolerating.