Oh well. You see, the GridFlow way has its advantages. By making patches out of multidimensional array operators, an incredible amount of work can be done with very little message-passing, and the supporting library (GridFlow) can do some insane vectorizing of operations. For high-bandwidth numeric computing, you shouldn't have zillions of objects passing around zillions of messages... it would be over 100 times slower (I tried)
I don't see the point. dyn~ does nothing else than employing the standard way of pd messaging and dsp processing. Thus, one can use all standard objects and even the highly praised Gridflow ones.
So it really depends on the case. I am conscious there isn't a single best way of doing things. The dyn~ object will have its niche too.
So be it.
greetings, Thomas