This is a continuation of my question here.
Page 635 of this book by Matthew Schwartz effectively says that the $\partial_\mu J^\mu_B\neq 0$ where $J^\mu_B$ is the baryon current i.e., the baryon number is not conserved at the quantum level. However, this term can be shown to be a total derivative, and hence, any Feynman diagram with this vertex will contain a factor $\sum p_\mu=0$. Since perturbation theory is based on Feynman diagrams, such a term cannot contribute at any order in perturbation theory.
However, the term $\partial_\mu J^\mu_B\neq 0$ can itself be calculated from triangle diagrams.
$\bullet$ Does it not mean that baryon number violation is possible (or at least calculable) even at the perturbative level?
$\bullet$ If it's a non-perturbative effect why is it computable using Feynman diagram?
$\bullet$ If I understand Schwartz's argument correctly, the anomaly is derivable from Feynman diagrams but the anomaly itself does not give rise to new Feynman diagrams. Is that right?