In research we don't measure "schizophrenia-ness", we measure people's experience, and so the data is still valid when the DSM goal posts move.

Let's think of some specific examples. Many research studies start with a question that they want to answer, such as "does this medication reduce the symptoms of depression?"; does this therapy reduce distress?" or "does having a diagnosis of schizophrenia cause stigma?" 

These are all questions which can be (and have been) investigated through empirical research. However, in all of these examples, although the groups of participants might be selected because they have a diagnosis of X disorder, the main things researchers would be looking at would be changes in scores on self-report questionnaires or interviews. These are independent of whether or not you have a diagnosis: I don't have a diagnosis of schizophrenia, but you could still give me a few questionnaires on "positive and negative symptoms" and get a score from me. So in essence, if I'm a clinician and I'm wondering "will this treatment/therapy make my service user feel better?" then I still have that evidence. In that sense, what the DSM decides constitutes a mental health problem is irrelevant.