Both methods have their disadvantages.
You can keep both dimensions in the same layout, in many games the performance impact will be minimal. However, you will have to keep these dimensions separated, for example on different layers. Also, you will need to disable everything for inactive dimension - collisions, solids, stop enemies etc. And you will have to remember filtering out inactive dimension in all events. For example, when picking enemies - only pick instances on a particular layer.
With switching layouts there is a problem of syncing. You'll need to give Persist behavior to all changeable objects, remember player's position etc. Also, if you have a lot of graphics, there will be a delay when switching to another layout, which you'll have to mask with some transition effect like fade in/ fade out.