I've worked in electronics manufacturing for 30 years. Every multi meter we have is sent to our calibration lab once a year for checking and adjustment. I don't have the exact numbers, but recalling from my memory, I believe about 5-10% of them would be out of tolerance and have to be adjusted.
Normally, the variation is fine for 120VAC or 12VDC measurements, but can be a big issue of for very small voltage, resistance, current or capacitance when we were working on circuit cards and components with very stringent tolerances.
By the way, the units that we used to check calibration are calibrated to National Institute of Standards and Technology (NIST) standards. These calibration tools can cost tens of thousands of $$ each.
As mentioned above, the Fluke meters were some of the most reliable we had, and rarely went out of calibration tolerance.