The existing test cases were created based on understanding the code. Creating good tests that cover every case you care about without understanding the code is not possible.
Ultimately there isn't enough information on how to most effectively maintain and evolve large, complex software with LLMs yet. We aren't even a year out from Opus 4.5.