standing violations under convention (ii): after detection the fruits are never used; the window before detection is charged by late debit

Maintainer decision, landed by the second follow-up. The per-block charge is restated under the adopted settlement convention: with the winning continuation's evaluation at least the window and any one counted violation, the loss against the bid is at least ℓ = ϖ − D + w with nonzero prices — the price term in the evaluation cancels the price term added back at settlement (BRIAFollowup2.price_cancels, standing_block_loss_ii, cross_block_blocked_ii); the residI form is the zero-price case. The headline: once a violation is detected and its taint recorded, use is decidable at choice time from the interface and compiles into the menu's structural count (nKnownWith), so a using option evaluates below inquiry by ranges alone and is never chosen (after_detection_never_used) — after detection the fruits are never used, not merely unprofitable. The window before detection: taint applies from commission, recomputed at detection over the record (detectAt, taint_only_from_commission); each window block is charged after the fact by a debit of ϖ against its winner's allowance — late debit equals late settlement (late_debit_eq_late_settlement, window_block_charged) — collected by the greedy schedule iff the cumulative allowance covers it (greedy_collects, collected_iff; the carry-over condition, with the uncollected part carried, uncollected_eq); what remains exposed is the window's length times the per-block advantage before the debit and only the uncollected debits after it (window_exposure). A violation never detected charges nothing: undetected commission is the count-integrity boundary (never_detected_never_charged), a different residual from untrackable fruit (knowledge, out of scope). Rejected alternative: charging detected use per block (the first follow-up's reading), which leaves it merely unprofitable where B.1 excludes it.