Karpova: The reason why this endgame deserves to be famous seems to have been forgotten, but it may be worth pointing it out again:
Rubinstein is faced with a new situation, a new problem, and now has to find a solution to it in OTB play, after a long and exhausting battle already. He comes up with an ingenious, and finally successful, solution to the problem. His solution proved not to be the best possible one. Furthermore, neither his nor Salwe's play in the endgame were faultless.
It may be worth adding here that Kmoch doesn't include this game in 'Rubinstein's Chess Masterpieces / 100 Selected Games', although he gives three games from Prague (1908). This indicates that it is no new discovery that play was not flawless.
But this doesn't distract from its importance: In a way, Rubinstein boldly presented a theory, a conjecture on how to win this special kind of endgame. Salwe was unable to refute it. And this endgame, this theory, provided later analysts with the model to research the endgame on their own. They understood that the more obvious mistakes didn't made the endgame study futile. And when digging deeper they found that Rubinstein's plan was not the best. Instead, they offered a better solution and enriched chess theory immeasurably. And Rubinstein's bold conjecture, his first offer of a solution, although far from perfect, provided them with the basis for their research and made it possible.
This is perhaps worth keeping in mind, when you can solve the endgame in the Nalimov tablebases (win in 48 after 60.Bg5) by simply inserting the position. Or when you mention the winning 57...g5, given on p. 140 of Donaldson & Minev, Volume 1: Uncrowned King, 2nd edition, Milford, CT, USA, 2006.