The Selection Bias is very difficult to check out in a pilot study. Only proportional representation of all groups can, in a way, indicate the absence of selection bias.
Interpretation bias and Non Response can be studied in a pilot run. Which questions are not answered? Can these non responses be explained?
A logistic regression of Non Responses with the rest of the questionnaire as explanatory can indicate whether the non response is 'at random' (i.e. ignorable) or if it is 'predictable' (i.e. non ignorable). In this case the Non Response will affect and biased the results of the study.