This seems to keep the important bit about Redis single threading, that is that you have exclusive access to the db in a script or `multi`. That property allows for a lot of things that wouldn't be possible otherwise. Also leads to very predictable performance characteristics.

That said, Redis has always been faster than I've needed, and that's leaning on it pretty heavily, so I kind of agree with antirez that adding this complexity may not come with much benefit. Will be interesting to see where this goes.