The old manual still applies because history has taught us not to confuse a laboratory exercise with a proven market solution.
A simple benchmark on 250 assets of the S&P 500 with a 64-qubit system is not proof of the viability of large-scale portfolio optimization for practical application.
We require performance evidence on much broader asset universes and in real market conditions, including transaction costs and liquidity constraints.
The 1997 Asian financial crisis showed us that models optimized on subsets of data can fail catastrophically when faced with the complexity of global markets.