Higher-scale testing using the Bluesky firehose?

[cassidyjames]

Discussed in the Apr 15 triage: we'd like a higher-scale testing framework that goes beyond current test coverage. Someone mentioned that the Bluesky firehose could be a good candidate for this; it's public, high-volume, and available via multiple atproto/bsky SDKs across languages.

Does this make sense to folks? Anyone want to take it on? :)

[ThisIsMissEm]

I think you could probably do a synthetic XRPC event stream for subscribeRepos which is [what the Relay is actually giving you, and then you wire that up with at-kafka (you just need to set the --relay-host to the synthetic data server.

You'd also need a plc directory running which actually stores the identities since at-kafka resolves all identities coming in.

cc @haileyok