I think it's worth elaborating on Sloth's post:
When an amp is operating in class "A", the current is allways running through the entire power amp. There are (generally) two types of power amps that can run in class "A" a push-pull amd and a single ended amp.
In a single ended amp there is only one output device (or a series of output devices in parallel) which is conected between a single power supply rail and ground, and is typically capacitively or inductively (transformer) couple to the load to remove the DC offset voltage. The offset voltage exists because there is only one supply rail so the output has a DC voltage on it somewhere about halfway between the supply rail and ground. In this configuration, the output device(s) are biased such that quiescent current (continuous current that flows even without a signal present) sets the operating point of the amplifier. When a signal is present at the input to the power amp the quiescent current is modulated producing a power amplified signal that drives the load. The advantage of this type of circuit is that the output device is allways running and ready to swing the drive to the load, so they are very responsive amps. The downside is that you have to AC couple the signal to the load through caps or transformers, and that the quality of the amplifier (distortion products) are directly related to the performance characteristics of the output device.
A class "B" amplifier is a design that puts into place two symetrical (hopefully) amplifiers, one between the positive rail and ground and one between the negative rail and ground. As the name implies, one half of the amp works on the positive side of the signal, pushing current through the load to ground and the other side pulling (sinking) current through the load from ground. A push-pull amplifier can be forced into class "A" operation by forcing current through both positive and negative side of the output amp. This current only flows in the amp and does not find its way to the load. The advantage of this type of amp is that it can be designed to compensate for some of the natural performance characteristics of the devices used. These amps are sometimes called "linear" amps because they are able to reduce non-linearities that result in distortion. ANother up side of these amps is that they naturally have an output that has no DC offset voltage, so they do not have to have DC blocking caps or resistors in the signal path. The downside of these amps---most especially when in class "B" mode---is there is a little zero crossing glitch when the signal goes through zero and on side turns off and the other side turns on. This is called "zero crossing distortion. (Our old descrete designed were of this type and did suffer from modest zero crossing distortion.) When you force a push-pull amp into class "A" operation, you get a region of operation where both plus and minus sides of the amp are both on, and you get a reduction of zero crossing distortion. Technically, this is called class "A/B" amps.
Now---and this is an important point---while you can force a push-pull amp further and further into class "A", you can't make a push-pull amp into a singal ended amp; and when knowledgable audiophiles discuss class "A" amps thay are mainly talking about single ended amps and not highly biased push-pull amps. This is because there is a school of thought that say that says a single device, while having higher distortion technically, also has a naturalness that can't be achived with a more complex design. Saddly, this issue is so complicated that the subtlety is lost on most, and marketers feel free to "steal" some of the thunder from single ended, truly class "A" designs by promoting highly biased push-pull designs as "Class A". Technically, there is nothing incorrect about this, and I am, of course, as guilty of this as anyone else. But with the rampant use of this buzz word I simple have to promote the new modules as class A to remain competative. I don't really feel too bad about this as the technique of forcing push-pull designs into class-A is benefitial.
The above discussion is a rather serious simplification. For example it is possible with multiple supply rails to design single ended designs that do not have to be coupled through caps or transformers. Output tranformerless (OTL) tube designes for example. And it is possible to design push-pull amps that are not simply class "B" topologies forced into class "A". The Diamond buffer-like design of the new modules as an example.
Keven, Tkam, or any of our other experts might want to chime in here to improve or correct my answer. This stuff isn't rocket science, but it heads in that direction. Simple statements like single ended is better than push-pull, or discrete is better than ICs, while possibly true in the extremely expensive end of things, are simply irrelevent in the face of the various cost compromises needed when designing affordable amps. While we have gone to a discrete design in these modules, I certainly don't rule out IC solutions as better and better chips come out. For example I think the new 8397 chip we've incorporated into the AirHead/BitHead is very sweet and it would be very hard to deliver it's level of performance at the same price with a discrete design.