Essentially, I think Hiho is saying that to adjust azimuth without altering VTA, the adjustment must be made downstream from the ~23 degree offset angle that is typically built into the geometry of fixed bearing tonearms. (It's not always 23 degrees; I own two tonearms, the Kenwood L07D tonearm and the Dynavector DV505 where that angle is less than 23 degrees.) If you adjust azimuth by twisting the arm upstream from the bend, then you are also changing VTA. Is that correct, Hiho?
Never having seen any of the tonearms Hiho mentions as exceptions, I don't readily understand how they work, unless the headshell itself twists about its own axis, as in the case of a removable headshell with built-in azimuth adjustment capability. The newer Reed tonearms also avoid this issue by allowing azimuth adjustment right at the headshell, but this adds several grams in terms of effective mass.