What it seems to say is that any of these may reduce polyphony:
-- some combinations of waves and osc mod on osc1
-- some waves in osc2 ????? is this serious or a typo ?????
-- filter routing settings
-- choice of filter type for filter 2
-- drive or waveshaping
-- vocoder modulator settings
Let's see what can I do without reducing polyphony?

Seriously, has any one gotten a clear answer from Korg or done experiments to know what settings will reduce the number of voices and by how much?