•We address scientific collaboration within Large Scale Research Infrastructures.•We identify four patterns of collaboration between instrument scientists and users.•Focus on the co-existence of multiple collaboration types within the same organisation.•We express partner dissimilarity in terms of expertise gap and co-development focus.•We suggest that LSRIs are important settings for collaboration on complex science. Over the past decades, Large Scale Research Infrastructures (LSRIs) have come to play a central role in providing scientist-users access to highly specialised scientific instrumentation and experimental conditions. Collaborations between (permanent) instrument scientists and users are at the core of these organisations, yet knowledge about the nature of such collaborations and their development over time is surprisingly scarce. In particular, we know very little about the interrelation between the individual and organisational drivers of collaboration. Based on a qualitative study of scientists and their collaborations at Institut Laue-Langevin, a world-leading neutron source, we identify four typical collaboration patterns, which reflect particular configurations of (dis)similarity between instrument scientists and users in terms of perceived expertise gap and co-development focus. Our findings suggest that the co-existence of multiple collaboration types within the same organisation plays an important role in the long-term success of LSRIs. In addition, we contend that dissimilarity can generate productive collaboration to the same extent as partner similarity; not only at the organisational level – co-existence of different types of collaborations across the LSRI, favouring the co-existence of a broad range of instrumentation – but also at the individual level – where instrument scientists benefit in terms of more productive collaborations over time despite the cost of learning involved.