> guarantee type stable allocations: once an address has been used to fulfill a request for a certain allocation class, it will only be used for that class. Slitter doesn’t overlay intrusive lists on top of freed allocations, so the data always reflects what the application last stored there. This means that double-frees and use-after-frees only affect the faulty allocation class. An application could even rely on read-after-free being benign to simplify non-blocking algorithms.

This is really fascinating. Does anyone know how much the typical cost of such types of guarded allocations are? is it in the single/double digit percentages?