Within Kirkwood theory, we study the translational diffusion coefficient of a single polymer chain in dilute solution, and focus on the small difference between the short-time Kirkwood value D(K) and the asymptotic long-time value D. We calculate this correction term by highly accurate large-scale Brownian dynamics simulations, and show that it is in perfect agreement with the rigorous variational result D<D(K), and with Fixman’s Green–Kubo formula, which is re-derived. This resolves the puzzle posed by earlier numerical results [Rey et al., Macromolecules 24, 4666 (1991)], which rather seemed to indicate D>D(K); the older data are shown to have insufficient statistical accuracy to resolve this question. We then discuss the Green–Kubo integrand in some detail. This function behaves very differently for pre-averaged versus fluctuating hydrodynamics, as shown for the initial value by analytical considerations corroborated by numerical results. We also present further numerical data on the chain’s statics and dynamics.