Snuffy was referring to the other axis of the scale, for balance beams. Your "standard" axis is the Z axis, for zeroing the scale. However, if it is also out of level in the X axis, it causes drag on parts of the scale that should not be touching.
I ran a fairly involved test a few years ago**, to see just how much I should trust my balance beam scale. With a 15 degree tilt (top of the scale leaned towards me, with the standard axis zeroed), the 5-0-5 was off by 4+ grains with a weight of 55.5 gr selected on the scale. (55.5 is arbitrary, but was selected so the "tens", "ones", and "tenths" were all used. However, 55.5 gr is a good average for the range of max loads in several cartridges I load.)
Most upsetting, was that the reading was not repeatable. I got different (bad) results with each attempt.
**Some of you may remember my post about the various accuracy tests and per-tenth-of-a-grain deflection of the RCBS 5-0-5, from 2.0 gr through 400 gr. I can't find the original post. I must have titled it oddly, and used different language in describing the tests.
I know most reloaders think that kind of testing is a complete waste of time. I did it, because I wanted to know how the scale reacted in different weight "regimes", and how it reacted to different variables. My background is in aerospace. Nearly every tool we used was calibrated on a monthly, quarterly, semi-annual, or annual basis. If a tool experienced shock, vibration, or contamination, it was recalibrated or replaced. Yet, reloaders assume their scales will remain perfectly calibrated for life.