Sure, it's a useful distinction for whether users need to care about safety but not whether the underlying code is safe itself, which is what I wrote about.

No or very little but verified unsafe internal code is the bar for many Rust reimplementations. It would also be what keeps the code memory safe.