Ah well, as mentioned, not a deal breaker by a long shot. Just strange...
Since Rigol uses an analog switch instead of a reed switch or relay, one possibility is that the ground offset comes from the analog switch having slightly different offset voltages on its ground and signal inputs. In that case, Rigol would need to provide a manual ground calibration option - either let people manually tell the scope how far out it is to set the offset or let the user initiate a manual auto-cal using the open/shorted inputs instead of the internally analog-switched ground.
Well, for one, the offset is already visible during the auto-cal. That is, even during that procedure the trace is never centered on the 0V line. And then, there is always an offset, regardless of the V/div setting (although it is slightly different depending on the V/div setting), and on all channels, so it's not just a matter of the auto-cal only adjusting at a specific V/dev setting correctly, then re-using that value so that other settings show it.
They could simply add another step to the auto-cal routine: for each V/div setting, and for each channel, take a meassurement with a sufficiently deep memory setting, average it out to get the offset, and store that. This should result in the traces being centered around the 0V centerline as good as possible.
Or as you said, let the user define a manual offset calibration for each V/div and channel setting.
I'm wondering if this issue is only present on the DS1000Z series, or if other model ranges from Rigol have it too.
Greetings,
Chris