Just a thought: use the centerline of the groove as your "curve". Now, instead of using the center of the pin as the "point", add a primatic joint between the center of the pin and the "point" frame, so the point that is following the curve and move with respect to the pin center. In another view, the pin can have a bit of play with respect to the curve. You can use the "limit" in the prismatic joint to specify the range of the play. You probably need to add a small mass to the point frame to avoid zero mass.
Another thought: use point cloud to represent the left and right side of the groove, and apply contact force modeling between the groove and the pin cylinder.
