In conservative (or complete) trace-based distributed fault diagnosis it is common to compute, for each component, a preliminary local estimate consisting of all possible traces that can generate the same local symptom (i.e., a collection of sequences of observable events) as the one being received from that component up to a specific time instant. Then communication among local trace estimators can be used to refine those preliminary local estimates. An approach has been proposed by Su and Wonham to perform such refinement in terms of achieving global or local consistency among local estimates. But that approach does not consider timing mismatch and channel distortion in the process of collecting local symptoms, which may result in incorrect estimates in practical applications. In this paper, we will provide a new approach to handle trace estimation under timing mismatch and channel distortion.