Purely greedy or not, there is some measure of "goal outcome" that was previously being solved for with the token selection function, and the goal was "complete this text with the best (surely, otherwise what are we doing?) next part, and sometimes the best next part is a little bit random just to keep things interesting"

Now the goal is either "identify the meaningless interesting bits and swap them out with 0% loss in the direction of the original goal," or "perturb some small selection of the output towards my secondary secret goal of watermarking the text."

It would be quite impressive if they managed to identify with 100% accuracy the tokens that "don't matter" and are free to swap with whatever signalling tokens encode the AI scarlet letter, but most likely they are not 100% accurate, and that means the output is worse off than without the watermarking logic.