As I understand it, Crema's fixed length loops gives it the computation power of primitive recursive functions. Which is enough for most programming applications. However, that assumes that a program is something which runs, computes a result and is done. A lot of programs fall outside that scope — even if they do not require more computational complexity per se.
A daemon or an operating system are typical examples of programs without bounded computation times. Not because of their computational complexity, but because they continuously (or by demand) produce new results. The interesting property for them is not termination, but productivity. They must keep being productive, and not become unresponsive. This is much more subtle notion than termination, but I think the programming language Agda has come some way with its co-recursive data types.
A daemon or an operating system are typical examples of programs without bounded computation times. Not because of their computational complexity, but because they continuously (or by demand) produce new results. The interesting property for them is not termination, but productivity. They must keep being productive, and not become unresponsive. This is much more subtle notion than termination, but I think the programming language Agda has come some way with its co-recursive data types.