It's nice if a language has a static check to prevent infinite loops when they're unnecessary, but this guarantee is less useful in practice if you're going to run the code.

A loop over a billion items (for example, due to a bad SQL statement) can hang your system indefinitely, just like an infinite loop. You still need timeouts, or a progress bar and a way to cancel. If running the code costs money, this is especially important.

A static termination guarantee is most useful in a proof language, when you're not going to run the code, just prove things by compiling it. That's when you really need it. You assume the function returns a value, and if there's any way it could fail, the proof is wrong. But performance is irrelevant if you're not going to run it.

For a configuration language, banning recursion might still make sense, though, since it's another way to write confusing code.