If you rerun the topology validation on the pre-fixed data (don't make any corrections or mark any exceptions), you can then export the topology results. This can be added to an mxd just like a feature class. You can symbolize by type of error, so you can compare this to your current topology.
However, you cannot select or intersect a topology error layer like a normal layer, but you will be able to see differences and overlaps/gaps. You could also look at your pre-fixed layers, which will be selectable and which can be spatially selected or joined.
But you have to keep in mind that the whole topology workflow is not designed for big changes like shifting most of a dataset. After edits like that, most of your data will be marked "dirty," because data has changed, which means it will be rechecked during validation even if it was at one point marked as an exception. So new topology errors will be flagged; you can definitely get more errors after such a validation because there may be some areas of exceptions that will not be re-checked and the new error shapes will not include those pieces. In other words, you can get more but smaller error shapes. If you revalidate your fixed data from scratch, you should in theory get the simplest, fewest errors. This is easier to understand if you symbolize the dirty areas, errors, and exceptions in a map.
When I have revalidated from scratch, I had had to bring in new rules/features. There is probably a less drastic way to do it, such as selecting all your exceptions and marking them as errors.