If you're attempting to use your measurement to help weed out "outlier" parts, even though they may meet specifications, then a gage R&R may need to be performed more often.
For example, the precision of your parts' process when you do your initial GRR is x. Now, it's 3 months later, and the part's process is now .5x; in other words, the process creating the parts is now 50% more precise than it was before. You may now not be able to discriminate the outlier parts anymore; you'll need to perform another GRR to see how capable your measurement is at discriminating between parts with much less variation than before.