Having determined a filter, say (0,0,-.5,-.5, 0) on the original data mesh, we can now interleave both t and x and expect to use the identical filter. Everyone knows how to interpolate data on the time mesh. The above filter interpolates space by extrapolating leftward. To extrapolate right, we simply reverse t and x in the star. So the spatial interpolation problem seems to be solved whenever adjoining traces are perfectly coherent, regardless of the signal shift. A ``spatial aliasing'' issue does not arise.
When the coherence of the channels is less than perfect we have an interesting question remaining, but it does not seem to be a critical one.