I'm in opposition to this for two reasons. 1. it is not a trivial change in IO, it is a very fundamental change in what can and can't be done. 2. It is not modular, programs written this way can only be combined as part of larger programs in limited ways.

1. With a normal form of output (e.g. 1 and 0) the set of problems that can be solved is the set of computable decision problems. With halting/non-halting output, the set of problems that can be solved is the set of limit computable decision problems. This includes famously non-computable problems such as the halting problem. This means there are certain algorithms with this output that cannot be translated to normal algorithms.

2. Since there's no way to do anything after producing a falsey result, there's no way to compose functions implemented with this output. You can't for example negate the output. I think programs should be modular and reusable.

And as such I think it should be up to the challenge writer to explicitly allow this form of output.

I don't think 2 matters for code golf. There is no goal to have reusable code.

On the other hand, i don't think solutions that require longer to compute as you can reasonable be expected of someone to wait for a result should be allowed by default. The problem is that is a very ambiguous and depend a lot on the computer, interpreter, ... However, it would eliminate the halting problem. But as i said, i was really not sure about this IO method and i am still unsure.

Maybe i should make the suggestion a bit different: A program may output a boolean value by runtime: If it finishes in less than N second (or minute or instructions) the value is true and if it does not finish in N+ϵ second the output is false. Where N is specified by the one posting the code and ϵ should be large enough to make the measurement easy. This would be easy to check and doesn't have the problem of forever.