FSM are anolgous to cellular automata, and since it is non linear algebrae, it is impossible to predict a finite sequence. FSM do indeed have most of the time regular basin of attraction with finite sequences, but there are chaotic evolutions that may results in finite (and longer than usual) sequences and infinite sequences without repetitions (given an infinite playground or CBP does the job).

And we put a lot of this chaotic systems in our software design.

The definition of a complex system is: a lot of simple system interacting with one an another. The physics and math of this typology of problem concluded that it is pretty much a non linear, non euclidean problem.

So far, what we know about these beasts is they are non predictible, BUT they are robusts around their equilibrum if not pertubated too much.

The stregnth of the beast is the network of connection but also its weakness.

If a certain one a perturbation happens, it can bring down everything (the butterlfy effect). It happens with a frequence and an odd we can't predict. And then it behaves in a way we can't predict.

So it is basically the ground of the internet: a non deterministic system used the same way a deterministic system is. What can go wrong?

We can't predict? Thus it will not happen anyway, so let's go back coding.