Hello,
I also see a problem that AC accuracy is off, but the other modes are fine.
It should be < 0.1% accurate in the ranges you described, even after many years.
Basically, the conversion is done by a TRMS chip, which has a fixed conversion factor, and the AC-attenuator resistors probably do not drift that much over time.
The DC-output from the TRMS chip (AD636 or so) is measured by the DCV part of the 34401A, and that does not drift that much either, especially as you found out, that DCV is still quite accurate.
You mentioned that you have no appropriate tools to calibrate ACV properly, so how would you know that you measure (compare) ACV properly?
As Guenthert already mentioned, there's the 'Geller-problem' with ACV on the 34401A, but also on many other DMMs with TRMS chip.
This simply means that their AC specification is valid only for signals between 5% and 120% of full scale, see footnote [4] in the specification.
If you measure consistently on 1% to 5% F.S. level, i.e. 0.5Vac on the 10V range, then you might already encounter systematically low readings (additional -0.1%) , worsening at even lower levels, and exact Null at I think < 0.1% F.S. signals. That your reading is 0.2% LOW, but not 0.2% HIGH, already indicates that this very probably explains the effect.
Make sure that you make your measurements at 100% F.S.
Distorted AC signals and similar might also be the root cause for worse accuracy, e.g. additional 0.15% for Crest Factor between 2-3.
If you can rule these effects out, then there might have happened an overload condition on the AC input attenuator network, i.e. especially on this 1MOhm divider resistors, R301, R302, see schematic, which led to an excessive drift. Maybe an overload with > 1kV occurred, which might also have damaged input circuit, C301, C302, U301. I would try to check these components.
Anyhow, if no damage can be found, but only a drift occurred, usually a new calibration will solve the problem.
Frank