Of course it goes into heat, but where? There is no shunt element, so what is dissipating the extra current that the load isnt using, especially during no load scenarios?
Maybe odd idea, but for dual tracking, how about instead using a differential i/o opamp for the error amplifier, running on something like +/-15v instead of +30v like the current error amp? One would, of course, need to invert the negative supply parts.