Topology in an versioned enterprise geodatabase is something of a mystery to me, as I've long held the belief that feature datasets within the egdb environment is looking for trouble. I still hold that belief, but have to deal with it none the less...
All that said, I am scripting a process that examines a topology and in doing so, exports the errors found.
In the production egdb, the actual errors themselves have been corrected in the life feature class, but the topology itself has not been (re)validated. When I run my script against the production egdb, I get no errors returned.
We also have a test egdb that mimics the production egdb. However, the line feature class has not had any edits made correcting the topological errors. When I point my script at the test environment, I get all the errors exported as expected.
Bottom line: when do topology errors disappear? I always figured you had to validate the topology to clear them out, but what I'm finding doesn't support that.