I think the wayland's devs have clear in mind that problem, so lets they work on the patch and through next versions to arrive to the right solution, as usual.
For example, the problem is already under discussion:
Actually I'm waiting for the PQ's work on subsurface and its refinement, because subsurface implementation will be useful for a lot of features and will help the completion of QtWayland in time for the release of Qt 5.1 I hope.For the issue with a client dying and leaving the compositor in a weird
resolution. Perhaps the API could change so that clients can specify
that the resolution change is contextual, and the previous resolution
would be set back when the client dies (it would look like the exclusive
part of the screenshooter).
Anyway it is really useful for testing mode_switch() callbacks.