Allow redispatching from within Reducers?

[Qata]

Should redispatching from within Reducers be allowed to prevent the double up of checking state with the middleware and then reacting to it with a redispatch?
In the current model, you're both reacting in a reducer to the state change, and preempting that state change in middleware before it even happens. This is far from ideal.