karalabe
commented
10 years ago Ok, the error was that the benchmarking code in the iris-go binding didn't wait for the published events to actually arrive, but rather disconnected immediately after publishing. This resulted in a gigantic set of events piling up in the relay, and getting delivered to the next attached connection.
During the pubsub throughput benchmarks there are a few weird things (a lot of out of bound events, seemingly looping deliveries). It would be worth an investigation.