The moment you watch a few seconds of a polarizing video, out of curiosity, they will start recommending more of the same. This is how people fall into more radicalized bubbles.
Only on Tiktok have I seen teen girls dancing along to clearly white supremacist content about racial superiority.
Anyone who thinks these issues are somehow unique or exclusive to Facebook is simply being dishonest. They are a byproduct of the anonymous, unfettered free speech we get from the internet not something intrinsic to recommendation algorithms.
They have allowed anti-vaxxer, conspiracy, misinformation etc communities to thrive and flourish and it's happened without recommendation algorithms. It happened because when people can see other's peoples comments they are able to follow them to discover more people like that.
On Reddit this is somewhat true, but it's a well-known part of Reddit culture (though this has been changing as they've been kicking out more subs) that each sub has its own set of cultural mores, opinionated mods, and memes. There's lots of drama/hate on subs from other subs but you'll see the same phenomenon on the Fediverse or, for that matter, in a neighborhood gossip group.
It's not like these views go away without algorithmic newsfeeds, it just that algorithmic feeds accelerate their spread.
On Facebook, I see potentially divisive posts from friends with very different political views. The difference with TikTok is that those friends don't post content.
I think if you explicitly seek out problematic content you can end in a rabbit hole on any large algorithmic app.