Let $f: M\to N$ be a surjective local diffeomorphism of noncompact smooth manifolds. Suppose that every smooth path is liftable, that is, for any smooth path $\gamma: [0,1]\to N$ and any point $p\in f^{-1}\gamma(0)$ there exists $\tilde \gamma: [0,1]\to M$, such that $\tilde \gamma(0)=p$ and $f\circ \tilde \gamma=\gamma$. By smooth path here I mean a map $[0,1]\to N$ that extends to a smooth map $(0-\varepsilon,1+\varepsilon)$.
Does it imply that $f$ is a covering?
In our situation being a covering is equivalent to the lifting property for continuous or rectifiable paths but the proofs that I am aware of (https://www.math.ucdavis.edu/~kapovich/EPR/cov.pdf) apparently do not go through when we know how to lift only smooth paths. The bottleneck is lifting homotopies.
Actually I am interested in cases where the class of liftable paths is even more restricted, though still abundant, like the class of piecewise linear paths.
In general, even if we can approximate uniformly a given path: $\gamma_i\to \gamma$ as ${i\to\infty}$, with $\gamma_i(0)=\gamma(0)$, $\gamma_i(1)=\gamma(1)$ and $lenght(\gamma_i)\to lenght(\gamma)$ by paths that lift to $\tilde\gamma_i$ with $\tilde\gamma_i(0)=p\in f^{-1} \gamma (0)$ and $\gamma$ is liftable on $[0,1)$, it does not allow us to conclude that $\gamma$ is liftable on the whole $[0,1]$. So, even if the answer to the question is 'yes', it likely doesn't follow for free by an approximation argument.