The terrifying decision for the API to return "*" on failure is a second-order consequence of memory unsafety, because it was a half-baked attempt to mitigate a common source of UB that people were encountering in the wild.

i'd say it is more a consequence of not having exceptions or other way of being explicit about error values.

That was a contributing factor, but there are plenty of other instances of C APIs that make use of in-band signaling like this. That the API was changed from returning one in-band signal to another is because the original value was causing programs to crash unpredictably due to UB.