JB, the argument is similar to finding eigenfunctions in QM. In this case the field must be continuous.

If the "current field" is "circular", then there must be a singularity somewhere. The line integral is then zero if the interior region of the contour does not contan that singularity.