In Shankar's QM book pg. 224, it was said that the contributions $Z=e^{iS[x(t)]/\hbar}$ add constructively near the classical path $x_{cl}(t)$ since the action $S[x_{}(t)]$ is stationary here.
As we move away from the classical path $x_{cl}(t)$, destructive interference sets in and the contributions $Z=e^{iS[x(t)]/\hbar}$ cancel each other out.
It was also stated that one can crudely say coherence is lost once the phase differs from the stationary value $S[x(t)]/\hbar \equiv S_{cl}/\hbar$ by about $\pi$.
How does this specific value of $\pi$ come about? Why can we say that if the phase difference is larger than $\pi$, destructive intereference occurs?