Due to the recent @Soup's post I have decided to do some housekeeping. I've been getting frustrated at lemmy's performance at times as well and this simply was a wake up call, if you will.
I am sorry about no advance downtime annoucement - Sunday is the only day I can really put any meaningful amount of time into lemmy.
Things done today:
Upgraded database VM 1 core 2GB -> 2 cores 4GB. Double the compute, double the memory.
Adjusted database config to account for increased resources
Adjusted hugepages config to account for increased database's requests
Updated both lemmy and database VMs
Rebooted the lot
Thank you for your patience. I will also use this moment of focus to write up a financial report in a separate post.
EDIT: I've also marked all instances known to lemmy.cafe as active. What this means is that lemmy.cafe will now keep retrying to federate to everything very aggressively. This is a compute-intensive process and will also impact performance for a few hours, until exponential backoff kicks in. I've done it to revive any falsely-marked-as-dead instances; there's no fix for it on lemmy itself.
I have seen a similar problem reported on sh.itjust.works matrix chat - in their case what helped was toggling the sticky post. I'll ask there to do that for this community. You might also want to reach out to mods directly - they might get to do it sooner.
Supposedly this is a known federation bug, but I haven't seen a bug report mentioned.