[mjn]

The highest-profile ones also provide publicity, which I think will be the last of their advantages to go (if it ever does). Even if we moved fully to a world where decentralized metrics (e.g. citation-based metrics) were the sole evaluation criterion, it would still be beneficial to publish in venues like Science and Nature, because they bring your article to the attention of many people (including journalists, who further spread it), which results in many more citations than you would get for the same paper published elsewhere. That's one reason, besides the prestige of the CV line itself, that people covet those kinds of publications: they're great for boosting your metrics.