The use of collaborative robots (cobots), designed to work alongside human operators in shared spaces, is predicted to increase in the coming years. However, successful deployment requires human trust, otherwise the collaboration may lack efficiency and human acceptance. As failures are an inevitable part of human-robot collaboration (HRC), it is important to examine potential failures, their likely negative impact on human trust, and how trust can be rebuilt. Interviews were conducted with 12 robot experts to understand their perspectives on factors affecting human trust, potential failures, and trust repair mechanisms in HRC. Qualitative content analysis was used and the results showed that failures can be classified according to their cause, symptoms and costs. This classification should be context and goal specific to enable optimal failure handling. Furthermore, the potential role of trust in third parties was uncovered and the importance of reliability and predictability for human trust in the cobot was confirmed. Experts mention both cobots and humans as being able to use trust repair mechanisms mainly following a pattern of failure detection, cause communication, failure correction, and communication of future steps.