for instance, having two material phases - material interface with strong discontinuity leads to jump in the stress field.
Nevertheless stress can be discontinous also for arbitrary FE space with piecewise linear functions because of the gradient of that function is constant over element and hence discontinous over elements.
For elements like truss and planes we have c0 continuity for displacements which leads to c-1 continuity for stresses. What kind of problem does this discontinuity bring to our model?
The stress aproximated in C-1 function space is less accurate than the C0 displacement field. If you measure the convergence of both fields in some suitable norm (L2) you see that the convergence of C0 displacement is of order better than convergence of C-1 stress field.