We describe ongoing work to increase trust in resilient software systems. Automated software repair techniques promise to increase system resiliency, allowing missions to continue in the face of software defects. While a number of program repair approaches have been proposed, the most scalable and applicable of those techniques can be the most difficult to trust. Using approximate solutions to the oracle problem, we consider three approaches by which trust can be re-established in a post-repair system. Each approach learns or infers a different form of partial model of correct behavior from pre-repair observations; post-repair systems are evaluated with respect to those models. We focus on partial oracles modeled from external execution signals, derived from similar code fragment behavior, and inferred from invariant relations over local variables. We believe these three approaches can provide an expanded assessment of trust in a repaired, resilient system.