Hi Dale,
This would not work because the cell constants are highly correlated with the other parameters in the model. You have to change the cell, refine all the other parameters, and then write down the R value. Many people have become confused by performing the test you describe. It shows a very steep dependency between the R value and the cell constants but the lowest R value is always at the values for the cell the model was refined with. If you perform your grid test and refine at each step you will see the broad minimum.
Dale Tronrud
realizing it, this is why I refixed my reply with "Not going into discussion whether this would be beneficial or not" -:) By suggesting this test I was thinking to catch some gross errors in unit cell parameters (that were probably not compensated by refinement of other parameters). My previous experience with some other similar searches tells me that you can find almost everything in this database. Pavel.