> Working software doesn’t always have to be correct or safe.

I feel like I am missing some very obvious point of yours because that statement in isolation I cannot agree with (I've read the rest of your comment and still can't find the extra context). Can you clarify?

Let me define the terms I used to see if that clears it up at all.

Working software: Does the software satisfy the use case?

Correct or safe: does the software enter into any invalid states? Are edge cases handled?

Does that help?