Hi Thomas,
That’s planned for a future GPlates release. It’s a bit more tricky than rigid raster reconstruction in terms of the way it’s implemented with graphics hardware acceleration (ie, the reason rigid raster reconstruction is fast enough to be interactive, which we’d like to retain for deforming rasters).
However, before that’s implemented in GPlates, it’s more likely that this will be done using Python libraries built on top of pyGPlates (see this post for more details on two libraries currently in development). That’s because the core approach taken in those libraries is to reconstruct points, and that will work quite well with the upcoming deforming functionality (in the next pyGPlates release). Essentially those libraries currently work by taking a uniform grid of points, sampling a present day raster (to give each point a z value), reconstructing those points to a past time and then creating a new raster using those reconstructed points and their z values. So the same approach can still apply when the points are deformed.