Digital research methods (e.g. online surveys and interviews) have recognised advantages when utilised within interactive research into sensitive topics, such as illicit drug use. These advantages include reaching a globally-distributed population and facilitating participant anonymity, often reducing risks for participant engagement. Digital communications can also be used to recruit members of hard-to-engage populations into research by providing the opportunity for deeper engagement and avoiding the 'hit-and-run' research style that decreases community trust. This paper describes our methodological journey with an easy-to-locate but difficult-to-engage population: people who bought illicit drugs from the online, darknet marketplace, Silk Road. The contribution of this research is in developing a robust methodology for conducting research in a socially and technically volatile online environment. Our approach builds upon an emerging body of literature exploring the darknet as a digital frontier for illicit exchange, community building and social activism. Over the course of nine months, we conducted 17 unstructured interviews through encrypted instant messaging (IM) with Silk Road drug buyers, whom we recruited through Twitter and Reddit (in the clearnet) and The Hub forums and IRC channel (in the darknet). A long project lead-in time was required to develop appropriate technical protocols to ensure complete anonymity of participants: including the encryption of email communication, use of Tor, familiarity with Tails, selecting an appropriate encrypted IM platform and data management practices. Days before the planned study launched, our research environment was seized by the FBI, necessitating new strategies for recruitment, and causing increased hostility and justified concern towards community outsiders among the target population. The community that had emerged around Silk Road fractured, and through our online presence, we were able to follow different threads of those networks to other clearnet and darknet sites in order to continue our research. Management of community responses to our presence in the fieldwork sites was important and involved consultation with community gatekeepers, dealing with trolls and threats to researcher safety, an official presence both in the clearnet and darknet, and presenting interim findings to the community for member checking. Collecting and analysing all deidentified correspondence data as well as screenshots of events that occurred during the fieldwork period proved integral for our interpretation of the final dataset. We see this kind of interactive research with darknet participants as a critical counterpoint to the passive analysis of digital trace data that is more commonly employed.