Hi Brian,
It's just a matter of technique. Craters can easily be eliminated by holding the arc at the end of the weld for a short time instead of quickly removing it. By holding the arc at the end of the weld, it allows the crater to fill up and helps reduce shrinkage forces.
If the welders are terminating their welds at the end of something such as a shear bar, they could start the weld at one end, weld towards the middle, terminate the weld about half way while briefly hold the arc at the end, then start at the opposite end, weld towards the first termination, burn through it, and briefly hold the arc to ensure the proper profile.
If there is a situation where craters are left and need to be filled after the fact, the craters need to be carefully checked for cracks before filling, because as you probably know, crater cracks can form due to improper arc termination. If no cracks exist, the crater can be filled by welding from slightly beyond the crater back into the crater and towards the weld.
brande indicated that welds with craters at the end cause the most problems. I definitely agree with that and since you are a CWI, is this where you're finding the biggest problem also? The older D1.1 codes used to state in what used to be Section 3 Workmanship, that "welds shall be terminated at the end of a joint in a manner that will ensure sound welds". I don't think this phrasing is used in the more current editions, but for obvious reasons should still be a general rule of thumb. Also, as you probably know, D1.1 has specific provisions for fillet weld terminations of nontubular statically and cyclically loaded structures.