Trouble with scan

The benefits of scan are well known. The intent here is to point out what kind of trouble we have with scan and BIST. The author then argues that functional testing methods might give a better quality if one can afford resources for manual test generation. Scan DFT is based on only two major fault models: stuck-at-fault and transition fault. Beyond these two fault models, the effectiveness of scan DFT is debatable. To achieve the kind of quality level we are looking for, more fault models will need to be applied in order to capture all types of defects. But even if we know how to create additional fault models (other than stuck-at and transition faults), how will we automatically generate test patterns for these additional models? And, how can any test generator or tester handle such a huge test data volume? How are we going to fault isolate these additional defect types?.