The migration from one database to another went well, though took about an hour longer than I expected. That's just processing time to dump and then reload a 500GB database.
Once it was complete and I tried to start the instance, I kept getting timeouts. It looks like the problem was with the web processes. I spent way too long debugging them. kept timing out. Nothing I did would fix it.
then I looked at the new database server. all 96 cores were pegged. wtf?
restart the database server, goes right back to being pegged. I stop the mastodon processes, and the load falls to zero.
then I wondered.... surely I wouldn't need to run an "analyze" command on a brand new database?
So I try it. kind of at my wits end. and sure enough the load drops to near zero and everything magically starts working.