I think you guys are all missing the point: they're talking about crazy high perfomances.
They've implemented it using non-locking algorithms, like LMAX's disruptor pattern (gigantic primitives arrays, CAS operation).
It's not a "joke" to be able to process 12 millions events per second on a single core in Java (that's where the LMAX disruptor patter is at right now, and it shall go higher once Oracle fixes the JVM so that it uses a faster CAS operation when possible [it's possible on many mainstream CPU and there are RFEs for this]).
The point here is that it is fast. As in "faster than anyone commenting here that is a joke could have ever build".