Consistency could be enforced more rigorously between the gewald parameter used in the creation of the table file (if kspace was turned on when it was created) and the currently-used gewald value. This could be done by adding a "gewald" keyword to the table file as Steve suggests, then having LAMMPS check that against its current value of gewald to make sure that they are consistent. If the keyword isn't in the file, a warning could be issued about possible inconsistency in the gewald parameter, which would allow backwards compatibility with old table files.
Paul