[dbaupp]

Data races can lead to other more conventional displays of memory unsafety. For instance, a read of a pointer length pair (a &[T] slice in Rust) while a write is occuring could resulting in a torn read where the pointer is the pre-write value but the length is the post-write one. If the original length was short than the new one, this can lead to a buffer overflow.

One framing of the MutexGuard problem is that the type wasn't declared in a way that reflected its semantics best, although it is clearly unfortunate that doing this is more complicated than the incorrect way.