The argument here is that if everything must be backed up by experiment before publication, then the number of ideas that can be generated and discussed in the scientific community will be throttled and progress will slow.
This is not an argument to be taken lightly. In a fast-paced field such as computer science, the number of ideas being discussed is obviously important. However, experimentation need not have an adverse effect; quite the contrary.
First, increasing the ratio of papers with meaningful validation has a good chance of actually accelerating progress: Questionable ideas will be weeded out more quickly and scientists will concentrate their energies on more promising approaches.
Second, I'm confident that good conceptual papers and papers formulating new hypotheses will continue to be valued by readers and will therefore get published. It should be understood that experimental testing of these hypotheses will come later.
So it is a matter of balance once more. Presently, non-theory research rarely moves beyond the assertive state, a state characterized by such weak justification as ``it seems intuitively obvious'', or ``it looks like a good idea'', or ``I tried it on a small example and it worked.'' Reaching a ground firmer than assertion is desirable.