Based on what I understand, those that test malware do it in a VM logging and redirecting all queries to external domains, in order to identify possible command and control hosts.

As a response, malware writers add checks for nonexistent domains. If, say, 5 domains known to be fake suddenly start replying, then the malware assumes that it's being executed inside a VM and stops doing anything, in order not to give researchers any clues. This malware just happened to check a single domain.