The reason that adding the capacitors have a large effect is because they performing what is commonly known as decoupling. Wire has resistance and, importantly inductance. As we should all know (simplifying it a bit) inductance resists the flow of current when the demand is a simple pulse or a continuous frequency. The several amps that your psu may be able to supply will not be available instantly due to the inductance and resistance of the wire. Even a shortish length will have some effect. By placing a large capacitor electrically and consequently physically close to the output transistors, makes more instantaneous current available.
As for a discreet design for the headphone amp. Should be possible. I will get my thinking cap on.
As for a discreet design for the headphone amp. Should be possible. I will get my thinking cap on.