[xwdv]

If you already know the optimal set of measurements why not skip the whole traveling back in time part and just change tomorrow today?

[gerad]

> The problem is that, only after the interaction does the metrologist learn which input would have been optimal.