Ok, so after digging deeper into available precision (0.1%) resistors I cooked up the following, even simpler solution:
This design slightly increases (from the previous solution) the resistances for the 6dB and 10dB attenuation settings, but simultaneously reduces the power dissipation in the divider resistors. Now 0.5W versions can be used (<250mW max dissipation). And only two resistor values are required 
The parts list is in the schematic. I will post pics of my prototype soon. This attenuator can be used for testing power amplifiers with up to 177VDC (or 250Vrms) on a single output power amplifier or if it’s a bridged one you can measure differentially 500Vrms.
Cheers,
Christian
PS: Is there a self-test mode available for the QA403 (with a pass/fail option). I have some doubts about the performance of unit I just received…