I think linearity will depend on a lot of things, most importantly supplied voltage and bias current (and likely input signal level as well). I would think it would also depend on the design of the amplifier and not only on what output device it uses.
In WA2EBY's article, he doesn't show or talk about any linearity measurements that he might have made, only that "No indications of instability, CW key clicks or distortion on SSB have been reported." which is not really a reliable test.
So I would either say as Jerry does and have a way to measure IMD distortion, or go with what the designer recommends. For SSB, unless you are really long-winded, is it probably is OK to have higher idle currents; 250mA at 20V is only an idle of 5W so not so bad if the heatsink is good for intermittant 50W. Regardless, it would be nice to see some test results showing IMD versus idle current so it could be set it as low as it can.