It depends on the extent to which your streamlines propagate into the GM. If basically all streamlines pass through that GM-WM interface into the GM, then the “reverse search” mechanism may well be the preferable approach. With the default radial search mechanism, if the maximal thickness of the GM were larger than the maximum search distance, then it may be possible for a streamline going all the way through the GM to not be assigned to a node. The counter-example is streamlines that terminate kind of at the GM-WM interface, but don’t quite intersect with the parcellation, in which case with the reverse search that streamline would not be assigned to that nearest node.
This is precisely the kind of non-standard use case I was thinking of when I implemented the multiple different mechanisms for streamline-to-node assignment.