If a scientists data was contaminated what is the best course of action

Of course, the best course of action is to re-analyze the samples again and generate a new dataset. This would be time consuming and may require further cost. But this would be the best possibility to eliminate bias from the data. In case, if re-analysis is not possible, then another option is to see the standard deviation among replicate. If only one replicate is contaminated, and other two are fine, then we can simply remove the contaminated sample considering it as an outlier.