We had a breakthrough! Synapse can now horizontally scale across multiple Python processes, with message sending no longer going via the main proc. This has spectacularly improved perf on over the last few weeks🎉 Read all about it:

@matrix Any chance to tell us how many workers and what exactly they're running on's server?
The docs specify all of those events you can offload but what exactly makes the biggest difference? Could we have some sane examples to base our configs on/cargo cult?

Sign in to participate in the conversation's Mastodon is one server in the network