There are many lesions that can be induced in brain and then not progress. The classic is the substantia nigra lesion in animals given large doses of MPTP or 6OHDA; once you kill that population of cells, you're pretty much done doing your damage. Granted, that comes with a major behavioral change, but it is easily imaginable that certain clusters of cells or cell types can be ablated in a region and then not have a progression and not show any obvious behavioral changes. The key to the last point is how hard were the looking at behavior....
Your questions regarding what types of patho show up in fixed/embedded specimens that don't show up in frozen tissue is right on the mark. It seems to me that, with care, anything you see in perfusion or postembedding will show in frozen tissue if processed right. Actually, I would expect that frozen specimens would show more...
Pathologists: please explain and venture a guess as to why the whole thing is dose dependent as well...