one of the amplifiers i've worked on using LEDs in the current sources was a Monster MPA-3250 amplifier. the amplifiers used a symmetrical design that was basically two identical but opposite polarity amplifiers in tandem. in other words, there were two diff amps, one using PNP transistors, and the other using NPN transistors. everything on the (+) side of the amp is duplicated with opposite polarity devices on the (-) side. the mirrored signal paths merge at the VAS (voltage amplifier stage). the VAS loads are current sources, and if you are building (or simulating in LTSpice) this is where things get interesting (and somewhat unpredictable) because there are two opposite polarity voltage amplifiers, and two opposite polarity current sources working directly against each other. if you look at just the two current sources in opposition, it can be an oscillator, especially when the voltage amplifiers, and their compensation caps are added. instead of a smooth "see-saw" action, it's more like a tug of war. in LTSpice, it can cause the kind of errors you see when you use an op amp and get the +/- inputs mixed up (all those wonderfully ambiguous "gm step too small" errors). this is one reason there aren't a lot of amplifiers made with a totally symmetrical input/VAS stage. i would need to look at the schematic for the Monster amp again to be sure, but i think the way they stabilized this beast was by providing a nested feedback loop from the VAS stage to the input stage, or by making the VAS current sources "mushy" (by adding some series resistance to the current source output, making them "nearly constant current").