In a tensorflow job I find it hard to imagine that a memory error would make a jot of difference... wouldn't the error simply be removed by the network training

In a tensorflow job I find it hard to imagine that a memory error would make a jot of difference... wouldn't the error simply be removed by the network training? If the error was in the execution then I imagine that the epoch would fail and it would then just be a matter of restarting.