It's not confidence inspiring when you are having issues like that - my RCBS version is a little bit wandery as well but nothing near the amount this one is. The answer would be to leave it on, checking and recording the drift amounts over a day at say 15min intervals and see exactly what it's doing. If it's really bad by documented results it's likely not good enough to be used for the job!
My process with the RCBS one is every 3 or 4 weighs chuck the calibration weight on it and rezero - it's more for my confidence than any error amount in the machine. With normal size cases, a 0.1grain amount either way won't affect much but then on a Hornet sized case 0.1 is a fair bit...
Bookmarks