My server was down for about 3 days, and when it spun back up, it looks like my timeline is lagged by those 3 days. It consistently catches up to the lagged backlog, so I suspect this is some offset calculation somewhere.
See if there are "incoming activities ready to process" messages in the file incoming_worker.log.
In my case, the issue was that I'd followed a couple of Lemmy communities, then unfollowed them, and the unfollow request didn't federate to the Lemmy instances properly (probably this: https://github.com/LemmyNet/lemmy/issues/2106). So I was getting a long backlog of boosts and such from pretty active threads. I ended up blocking the communities themselves - I don't think that'll have an impact on my ability to reply to a post or comment, since those show up as posts from the OPs themselves and are just boosted by the community.
~Isaac Schemm email@example.com writes:
> See if there are "incoming activities ready to process" messages > in the file incoming_worker.log.
Yep! That was it - 70k+ entries, and I now see why my VM (which is tiny) was thrashing the disk.
I've dropped all the entries in the incoming_activities table (don't really care so much for lost inbound) and crossed my fingers.
Have a backup, in case that broke things :-)