Cascode the active loads. This is the single most important thing. One more mosfet and gate stopper each. Can't over-emphasize this one. Might be best to design to put it in later, just for effect. Battery bias on the CCS is an improvement as well, but is a bit more complex and perhaps can be done o a rev. 2 ( and more comfort has been achieved with the design ).

I don't quite agree with the cheap leads to ForePlay. The original as drawn will do quite well( better I think by a mile, but that's opinion for the time being ).

regards,
Douglas