The question is in the title. Can a Poincare invariant Lagrangian lead to a path integral that is not Lorentz or Poincare invariant? If so, can I have an example?
A related confusion: on page 426 of Weinberg I, he seems to show that this is not the case, at least for the translation subgroup. However, on the next page, he uses the same argument to show that electric charge is always conserved, even non-perturbatively. Now this is not always true. Is it the case that he is actually showing it implicitly only to all orders in perturbation theory (but not at finite coupling), and that the caveat is that the free states do not necessarily span the Hilbert space?