Well, apart of these ontological questions we always could listen to Feyman and go with the "don´t think, calculate" premise. But, can we?
The fundamental calculational tool in string theory is the polyakov path integral. If you read the correponding chpaters in the string books they aregue that one virtue of string theory if that you don´t need so many feyman diagrams and taht you basically need one kind of vertex, the one in wich an incoming string separates in two outgoing ones. By a lorentz transformation that vertex is shown to be equivalente to ones in wich you have two incoming strings who join in a single one.
In QFT (sdecond quantized theory) you get a prescritpion on how the Vertex are form teh form of the lagrangian. In perturbative string theory they are put "by hand".
My question, of course is, why no other diagrams?. For example you could have a diagram with tow incoming and two outgoing strings, or an string breaking in more than two pieces. In fact you could, as far as i see, have an string breaking itself in an arbitray n of strings, and i don´t see that you could reduce these case to the simple one.
But i recognice that these could easily be a missunderstanding on the polyakov integral, may be someone could explain me if it is so.