MARSSIM discourages double sampling. The DQO process, which MARSSIM uses, explicitly sets objectives for both the Type II error rate and the retrospective power during the design process. Adequate initial sampling to achieve the desired power makes decisions based on the data more objective and defensible. A better solution to the issue of double sampling is to plan for data collection in two stages, and design the final status survey accordingly.
If your regulator allows double sampling, it should be decided upon during the DQO process. You should both agree upon the number of samples allowed in the second set of samples, because the Type I error rate could be as much as double the error rate for a single set of samples.
Class 2 or Class 3 survey units are not appropriate for double sampling. Concentrations in these survey units should not exceed the DCGLW, and should always pass with the first set of samples. The need for a second set of samples in Class 2 or Class 3 survey units raises the issue of misclassification. Double sampling is also generally not appropriate for Class 1 survey units having confirmed areas of elevated activity, or "hot spots."