It really depends on what exactly you are trying to do.
If you are looking to track the changes between the relationship of different type-2 dimensions, then a (fact-less) periodic snapshot table will be sufficient. If the changes are too rapid or your dimensions very large, you could look at temporal fact tables as a possible solution.