nobody needs to move to another platform. both lemmy and piefed show the same content, think of it more like using a different client that also has different features. both lemmy.world and piefed.world will continue to exist.
piefed does have some social auth support, which is currently also being worked on, but lemmy is not an auth provider that can be used with that. once social auth in piefed becomes more stable we will consider enabling it for supported providers.
i think the graph might not be fully accurate for the current hour? looking at it now it doesn't show that significant of a spike anymore. i don't see anything in our logs about federation issues from LW to p.d in the last 15 days.
there were some caching issues in lemmy-ui where it would unnecessarily eat up disk space for caching without even making use of it properly. there was a change done in 0.19.12 that was supposed to mitigate this, but for users who have already collected this it won't automatically delete the unnecessary cache until they logout: https://github.com/LemmyNet/lemmy-ui/pull/3150
even when clearing this manually though i still see this take longer than you'd expect to load, it seems that the image cache is still slowing things down.
i also had some delays on images on the front page before all media loaded. i was able to speed things up again by executing await window.caches.delete("image-cache") in my browser dev tools console, but that is certainly not something to expect from regular users.
where are you located? do you have an example of things loading slowly? for me things are loading instantly, but if you're not within the EU you're likely dealing with latency across the globe
yes, post_read marks which posts a user has marked as read. it links post ids with user ids and adds a timestamp on top to allow for sorting.
edit:
for comments, lemmy only stores the number of read comments per post, which is what goes into person_post_aggregates. this is a tradeoff, it has some limitations, e.g. when comments are deleted or removed, which affects the total count. as there is also a timestamp attached it might be possible to use that in combination with comment creation times, though this would likely impact query performance quite a bit more.
can you think of any solution that is not a variation of “keep finding fresh volunteers to work until they burn out”?
how would paying admins prevent burnout? the only difference i see here is that it is probably easier to find people willing to do it as a paid job than volunteers, but they can both burn out. this would just change it from "keep finding fresh volunteers" to "keep finding fresh job applicants".
I'm personally only involved in Ruud's side of things (mostly .world instances). Stux' platforms are managed separately, I can't say too much about those. Afaik finances between Ruud's instances and Stux' instances are also separate.
On the .world side, we currently have 6 active members for infra. For moderation, LW currently has 4 active instance admins plus some community team members with elevated privileges. Other .world platforms have moderation separate from LW. We certainly don't have resources to hire professional admins, but I'm sure that we would find a viable solution if Ruud ever wanted to leave things behind. Not all solutions require paying someone a salary for it, which seems to be your implication here.
I don't think "too big to fail" is as much of a factor here as the fact that LW is not the only FHF platform. Fedihosting Foundation, the non-profit behind Lemmy.World and our other platforms, existed before Lemmy.World already. While the Lemmy moderation team is working mostly independently from the rest of FHF, if the LW admin team disappeared there would still be FHF in a position to search for new admins and probably also at least temporarily step in without requiring to shut down the instance.
what kind of error did you see and what did you click on? a link to a post?