I actually find these ad-hoc size checks rather unfortunate. If we're actually out of memory then malloc should just return NULL and everything should be fine. The fuzzer's AddressSanitizer really shouldn't be crashing just because malloc would return NULL. I've had this fight many times with the fuzzer people, unfortunately. In reality these checks and the fuzzer ending early is actually decreasing coverage of the failure paths.

Thanks for the fix! And: well... I guess we have to live with that.