Agreed that's a concern.

There could be some visual language for how recently changes happened to the LLM-generated tests (or code for TDD mode).. then you'd be able to see that a test failed and was changed recently. Would that help?