I have an account on feddit.uk and lemmy.world with almost identical subscriptions.
When I visit the feddit.uk homepage and look at posts from my subscriptions sorted by “hot”, the two posts immediately after the two pinned Feddit UK posts at the top are 6 days old. The behaviour with “active” is similar. Posts shown are quite old, and neither hot nor active.
When I visit the lemmy.world homepage and look at the “hot” posts from my subscriptions, the first 10 are all with the last couple of hours. Active shows posts in the last 24 hours with a healthly number of comments.
Both servers are running 0.17.4
Basically, right now feddit.uk is almost unusable for me due to both hot and active being populated by stale posts. I need to search by “new” to get anything approaching useful - but when I do that there is just too much to sift through.
I’m not really sure what’s going on but this the algorithm Lemmy uses in theory: https://join-lemmy.org/docs/contributors/07-ranking-algo.html
Could it be something to do with lemmy.world’s federation skewing the votes/comments? The other thing is whether the ranking is cached in some way and that’s broken or not invalidating early enough. Don’t know.
I reckon something is just broken (at least for me). This post about lettuce has been my top hot post for days both on the web and via two different apps…
Just taking a cursory look at the code, it appears the ranking of posts per sort is stored in the database routinely via a background process on the server: https://github.com/LemmyNet/lemmy/blob/main/src/scheduled_tasks.rs#L31C10-L31C10
So, could be to do with that. If the server is underprovisioned, maybe it’s taking an overly long time to run each calculation and everything is being held up. Complete speculation, though, at this point. 🤷
Edit: this is actually the 1.8 version of the code. Prior to this it looks like this was being done differently (see: https://github.com/LemmyNet/lemmy/pull/2952/files). So like Emperor says, it looks some optimization work was done here as presumably the queries were running too slowly.