Why is lattice QCD called non-perturbative? Like, if you are approximating a smooth structure with a discrete lattice, isn't this like a perturbation from smooth space-time?
If Feynman diagrams are a perturbative method, why are Feynamn diagrams on a lattice/grid called non-perturbative?
 A: In general, by a perturbative approach, we mean an approximation of the form,
$$f = f_0 + \epsilon f_1 + \epsilon^2 f_2 + \dots$$
where $\epsilon$ is the perturbation parameter, for some solution $f$. That is to say, one can approximate the behaviour of the solution by this series.
However, summing all the terms does not mean you will recover the exact solution; in most cases perturbative series are asymptotic series.
On the other hand, lattice QCD is an approach which is not described as perturbation theory because it does not follow this scheme, and in principle one recovers the exact solution in the appropriate limit.
To convince yourself of the distinction, consider the differential equation,
$$\frac{\mathrm d f}{\mathrm dx} = g(x).$$
If we choose to discretize it (very naively), we obtain a linear system,
$$\frac{f_{i+1}-f_{i}}{\Delta x} = g_i$$
which is a totally different approach to plugging in a series expansion like the one above. That being said, Feynman diagrams are always a perturbative approach, so Feynman diagrams on a lattice are a perturbative approach, but putting the theory on a lattice is not what makes it perturbative.
