Although it's not the case here, I quite dislike whenever this article is brought up because usually it's used to shut up people that says Spotify is "not random enough", ignoring that
1. Regardless of intention, the random is subjectively broken according to a lot of users (Including myself, 3 songs in my playlist become my top 5 listened song solely due to it always "randomly" selected to the top.)
2. This article is old af and may not even be relevant anymore.
Although saying 'our users [are all wrong and stupid and have] fallen victims to Gambler’s fallacy' (obviously my addition in square brackets) is a bad look, the rest of the article details how they adapted the algorithm to make it seem more random even though actually it is less random.
There's no winners here, the randomness purists will say 'well if shuffle plays all the songs in alphabetical order, or all your Metallica songs first, or some other order, they're all perfectly legal permutations and as likely as each other'.
Then the majority, folks who just want to listen to a variety of music, who will say 'how can it possibly be random if all the Metallica songs are together or Pantera songs are together, or Christina Aguilera songs are together?'
I think in the end Spotify did the right thing here by opting to be less mathematically purist, and doing what users actually want from a shuffle.
There is a separate (and maybe unrelated) issue which is that dynamically generated playlists seem to heavily favour well-known songs, which may be what you're referring to.
But yeah, this article is coming up on ten years old now. I would be surprised if the shuffle algorithm is still the same.