Querying invalid geometries
We jumped from 8i to 10gR2 and didn't have too many problems.
We are now considering moving to 11gR2 and are considering the impacts of the strongly recommended approach to not have spatial data at a resolution higher than 5cm. Our data is a higher resolution than that and consequently we have many (many many) invalid geometries if we set the layer up with that resolution (tolerance). However, with the little bit of testing we have done so far, invalid geometries are being returned by queries.