Number of plays for a track is not that useful to determine popularity of a track. I prefer using a combination of "number of comments" a high Like-to-Play ratio (ex. 100+ likes for 1000 plays), repeat listens, and reposts. The people "gaming" the system by boosting their play counts where the play is merely a click (likely by a bot) rather than someone who listened to the song, take away from the value of this stat.
As a publisher and a listener, I would prefer to see Total Minutes Played per Track and see the play count de-valued or even removed. As a listener, the new stat would better indicate what other people are actually listening to (and presumedly like). For the publisher, this means a more level playing fields with those that promote "clicks" so they do not have an automatic advantage. Songs promoted by Soundcloud, should also consider this new stat.
There are obvious ramification to this change because Soundcloud would have to deal with the "gamers" in a different way since they would "adapt". In marketing research, we use algorithms to detect "speeders", those completing surveys faster than possible with a level of accuracy/honesty. In Soundcloud terms, this could be detecting and penalizing any account that is listening to more than is humanly possible, both in number of tracks and time.