The fact that is working as designed has no bearing on the question of whether it is a physics problem. You could just as well say (as Intel's press release seemed to to be trying to imply) that there isn't a fault at all, whether physical or not, because it is working as designed.
But your own description of it working as designed is within the context of a state-machine model, and in that model, there is no attack vector; that only appears when you take the physics of the implementation into account. There wouldn't even be a cache except for the temporal physics of memory access. That it works as designed is beside the point because the design did not take the physics into account.
Looking at it from the other direction, how do you conclude that rowhammer does exploit a physical phenomenon, without being inconsistent about what that means?
But your own description of it working as designed is within the context of a state-machine model, and in that model, there is no attack vector; that only appears when you take the physics of the implementation into account. There wouldn't even be a cache except for the temporal physics of memory access. That it works as designed is beside the point because the design did not take the physics into account.
Looking at it from the other direction, how do you conclude that rowhammer does exploit a physical phenomenon, without being inconsistent about what that means?