Is it true that for a symmetric TSP instance in the sequence of edges generated by successively:
- calculating the optimal 2-factor
- adding cardinality constraints on the edgesets of the 2-factor's connected components
an edge that constituted to the optimal 2-factor of the $i$-th iteration but not to the 2-factor of the $i+1$-st iteration can't reappear in any of the 2-factors of later iterations?
