Also you are forced to message-pass, which adds overhead of a processor talking to itself; for some applications you want shared memory and parallel computing rather than message-passing and concurrent computing.
You also might find that its behavior running on multiple nodes isn't optimal, because it's a generic solution; you might do better with a specific solution with a truncated feature set, written in a language with a faster runtime (SBCL; C++; Haskell; etc.). You might also want guarantees of order, execution time, or determinacy -- something else it doesn't give you.
Furthermore, you might find that it's not scalable on the programmer-end; more people may have experience with Python, Ruby, or Lisp. Or it may lack libraries you want, and thus not "scale" to cover certain types of applications, given the constraint of human developer time.
Trivially, Erlang; because the language and implementation, including Mnesia, go together.
However, you might find its performance characteristics don't suit your problem domain: it's not fast compared to C:
http://shootout.alioth.debian.org/gp4sandbox/benchmark.php?test=all&lang=hipe&lang2=gcc
Also you are forced to message-pass, which adds overhead of a processor talking to itself; for some applications you want shared memory and parallel computing rather than message-passing and concurrent computing.
You also might find that its behavior running on multiple nodes isn't optimal, because it's a generic solution; you might do better with a specific solution with a truncated feature set, written in a language with a faster runtime (SBCL; C++; Haskell; etc.). You might also want guarantees of order, execution time, or determinacy -- something else it doesn't give you.
Furthermore, you might find that it's not scalable on the programmer-end; more people may have experience with Python, Ruby, or Lisp. Or it may lack libraries you want, and thus not "scale" to cover certain types of applications, given the constraint of human developer time.
So the answer is, "it depends".