# Independent trial-balance calculations The 62 fixture histories and literal expectations were authored before the trial implementation. The authoring script imports no algorithm runtime. Expected status, account values, debit/credit totals, mismatch counts, source positions and calendar boundaries are declared independently. In the distinct CB-CLOSE-2026 branch, accepted opening contributes cash raw +100000 and opening equity raw −100000. September accrual contributes operating expense +25000 and payable −25000. Three additional USD accounts remain explicit zero. Native EUR cash/equity entries are not added. September ending columns: debit 100000+25000=125000; credit 100000+25000=125000. Raw residual is zero. Payable raw −25000 displays normal +25000 and belongs in the credit column. The October reversal debits payable 25000 and credits operating expense 25000. It is known at the September query K but booked October, so September stays 125000. For October cumulative reporting, expense/payable are zero and columns each 100000. Selected USD debit and credit leg volumes are each 100000+25000+25000=150000, which is not a closing balance. An authorized prior-effective adjustment booked October adds 1000 expense/payable only to October: columns 101000, selected debit/credit volume 151000. Economic September date does not silently move its booking month. Reported cash +1 makes reported columns 125001 versus 125000, residual +1. Reported cash +100 and opening equity −100 preserves equality at 125100, but differences +100 and −100 fail two account projections. Negative cash example: another debit other expense 150000 / credit cash 150000 yields cash −50000, other expense +150000, operating expense +25000. Debit ending balances 175000 equal credit ending balances 50000+100000+25000=175000. Cash normal is −50000, an abnormal-side diagnostic rather than an automatic arithmetic error. The atomic estimate correction retains original 25000, exact inverse 25000 and replacement 24000. September balance columns become 124000, while selected debit/credit volume is 100000+25000+25000+24000=174000. Before the complete correction receipt both members remain unavailable; at equality both enter. Historical p2/h4 and p3/h5 cases distinguish later complete validation coverage from the whole-group position visible at K. Report publication and visible position are authored independently; the expected report is never calculated from implementation output. A current period-close adapter separately requires p==frozen h. The wrong-economic-account case changes both accepted journal and reported vector consistently. It passes projection arithmetic and explicitly demonstrates the algorithm's inability to determine correct economic coding from balance equality.