This one is a tricky one. I understand the intent, but I think platforms need to shy away from trying to optimize this in particular.
One reason is you can't define "high-quality content". Thinking in terms of Tumblr's userbase, there are users who have very specific things they want to see, and if you try to focus on some measure of "quality" (how? engagement? bad idea!) you may just frustrate and alienate them by making it hard to do this curation. That means it's important to make sure you DON'T force an algorithmic feed, even if you're not doing the Threads thing and making it basically a giant sponsored post section in the timeline.
Another reason is that's not what people actually like about the internet. The internet is fun because it's a mixed bag: you get a lot of everything. Seeing only the best things every day is neat, but after a while you just get numb to it. What you want is something enticing and interesting. A good example here is YouTube in say, 2008 or so. That was crazy. There was not much of an "algorithm" to speak of: your homepage was just surfacing your followers and their interactions with others. This made a natural social graph of content for you to explore, and you could find stuff that people you like thought was interesting. The experience of crawling through this web of content is enticing in a difficult to explain way, but it would certainly be made worse if you just select stuff that an ML algorithm thinks I would like. Case-in-point, YouTube's recommendations for me are horrible, because it more often than not finds superficially relevant content that doesn't actually scratch the same itch as the stuff I do like.
Worse, I think over time having systems like this cause lots of feedback loops. People discover ways to game the algorithm and suddenly everybody is playing the game. Content that is de-prioritized by the algorithm becomes less visible, and makes discoverability of things that people would like to see much harder.
I suspect most people who sign up for Tumblr have a specific entrypoint: another person or account that they came there for. The focus on discoverability should, in my opinion, focus on moving outward from there rather than trying to push a bunch of viral content on the user in hopes that it will appeal to them. Same goes for a lot of other social networks that do literally the same thing.
It will help your metrics in the short term, but you may very well bleed even more dedicated users. Metrics can obscure a lot of important things...