These are two different kinds of reliability (test-retest reliability vs. internal reliability). Ideally, you would assess both - but if given the choice between the two, internal reliability is more critical. Of course, you don't need to use the split-half approach. There are other alternatives.