@BeAware @philip

Subscribing to the Flipboard relay with 4 CPU cores and 6 GB RAM, even with jvm memory set at 1gb, on the same server as my Mastodon instance, brought my ElasticSearch down at around 1pm every other day and crashed Redis also. It's a busy relay. But, that's on me for not having enough resources. (I think this is my last edit. lol)

@torspedia @philip Probably the simplest way is to subscribe to a bunch of federation relays but not sure if that will give any information to when it breaks organically.

Besides that, it's really hard to replicate server load because it may work fine one day with a number of users/followers/following and then it may slow down or stop working the next. Nothing changed internally, just some mass load from external instances causing the extra federation.