@sjw The hourly hentai bot stampede¹, for one thing. (And my disk usage several Postgres DBs on one box.) So until the management interface is up and running, I don't want to put these instances on a relay with bigger instances, especially ones with a lot of bots. Once it is up and the people running the instances have easier control over the relays, it'll be different, but I don't want to flood them before they can even decide if they want that. Even then, it'll be kind of a pain until the object proxy and refetcher is live.

¹ FSE has it, too, all these bots that post at hh:00, flooding TWKN with a million damn botposts in a row because everyone decided "Post once an hour" means they should post immediately at the top of every hour. I keep begging everyone to offset the bots so there's no stampede because it's a shit-ton of scrolling if you look at TWKN at the top of the hour. (You might have noticed lainbot posts at 4 minutes after the top of the hour, limerickbot at 13 minutes after. I pick random offsets for all of them, even the unlisted bots like pbort and jojobot, because the bot flood is the second worst thing to happen to TWKN, the worst thing being Gab.)