After some more probing, the problem manifests itself around U410 (a TL074 op-amp).
The output of U409 (the dynamic flatness correction DAC) is stable, goes into U410A and out comes the voltage jump (37,5 mVdc at that stage) which then moves further down through U410B, U314 (another TL074 op-amp), and Q401 to end up on the DACOUT lines, at least that’s what it looks like to me.
I’m still a bit sceptical, though, towards U410 being the culprit. This is quite strange behaviour for an op-amp. Also, the problem only shows up on sine wave output, not on square and triangle waves.
But if it didn’t come from U410, the only other option I can see is U407 (the main waveform DAC) somehow backfiring into the amplitude adjustment circuitry, which seems quite unlikely, looking at the schematic.
Any thoughts on this? I’ve attached the section of the schematic with my measurements for reference.
Edit: I replaced U410, problem remained. I recalibrated it two more times and got it to barely stay within the specified limits*, still just because of the single jump, otherwise it’s really flat. Given I don’t really have adequate equipment for measuring high frequency AC at hand, I’ll just leave it at this for now. That’s actually good enough for my use cases.
I’m still curious, though, what causes this weird jump …
* Depending on which device I use … the 34401A reads a delta of 0.0694 Vrms from 1 to 100 kHz, my 54645D oscilloscope reads 0.021 Vrms. Specified is a maximal error of +/- 0.03 Vrms.