Ironclad is capable of answering to out of memory conditions as well as hardware memory errors.
For OOM failures, Ironclad will try to free internal storage used for cache and retry allocation until it is deemed the situation is not recoverable, in which case it will panic. To this end, Ironclad will kill high memory consuming processes, this behaviour is configurable in failure_policy.
For hardware memory failures, like RAM sticks going bad, handling will depend on the underlying architecture. Behaviour is configurable as well under failure_policy.