Here's something else to think about. It was presented to me in a way that made it very easy to understand by Bob Carver. I read this in the "white papers" for the Sunfire amps and it makes good sense.
A transistor is a WATTAGE device, not just a voltage or current device. To make things as simple as possible, wattage is the voltage multiplied by the current ( amps ) in the circuit. In other words, a transistor may be capable of 200 watts of output. If the circuit is running at 100 volts, that transistor can only pass 2 amps of current to achieve rated output at spec. After all, wattage is volts x amps and 100 volts x 2 amps would give us the 200 watt rating. We'll consider this the operation within an 8 ohm circuit.
While running more transistors in parallel will increase the amount of current that we can safely pass due to sharing the load, we are still limited by how much voltage we can safely deal with. That is, if we want to enjoy an extended lifespan from the output device and keep it operating within the designed specs. There are drawbacks to running multiple devices to share the load, as you know have to worry about them all working together and doing so at the exact same time and rate. This is where "matched parts" come into play. Only question is, how well does one have to match the parts before "perfection" is achieved in such a situation ?
Now, If we took that same example circuit from above and went from an 8 ohm load to a 4 ohm load, the same device would now be trying to pass 4 amps of current. Since resistance is halved, current is doubled. The problem with this is that we are still limited to that same 200 watts of output with good linearity from the transistor, so something has to give. That "something" is voltage capacity. Since 200 watts divided by the 4 amps that we are trying to pass leaves us at 50 volts, we've now lost a LOT of headroom ( 50% reduction or -3 dB's ) in the amplifier. It is at this point that compression ( slight squashing of the peaks ) or clipping ( complete "smashing" of the peaks ) can set in depending on how hard we are driving the amp. Not only that, we've also increased thermal losses which are attributed to reduced amplifier efficiency. This loss of efficiency is evidenced by the increased amount of heat dissipated by the amp. You pass more power, you generate more heat. More heat causes the outputs to operate out of their optimum range, causing them to perform poorer and efficiency to drop even more. Kind of a vicious circle but that is what we are dealing with at this point in time with the limited technology that we have. Superconductivity tries to deal with this by "freezing" the parts so that thermal losses don't come into play. That is a whole 'nother ball of wax though.
While some of this is "speculation", my experience is that a transistor works and sounds MUCH better when kept relatively cool or at one would call "normal operating temperature". Going above that point typically introduces distortion, smearing and "grain". Since the laws of physics dictate that a lower impedance will pull more current no matter what, we're already fighting an up-hill battle. While some semiconductors are better / worse than others in this area, they will all run into the same problem sooner or later. Increased heat and increased amplitude almost always go hand in hand with reduced linearity.
In order to keep this from happening, you either need REALLY big heatsinks or some type of forced cooling system. The cooling system itself can become another source of noise and interference, so most designers try to avoid them if possible. Since heat rises, one of the simplest things that you can do is allow a LOT of room above and around the amp to achieve maximum air flow and natural cooling to take place. This will never hurt and can only help. People using amps "shoved" into a enclosed rack are at an instant disadvantage compared to those that have open sided racks with more room between shelves. Running an amp on a completely open "amp stand" and elevating the amp via cones, spikes, small wood blocks, etc... so that you can get airflow on all 6 sides of the chassis' is about the best that anyone can do.
The bottom line is that a lower impedance speaker presents an entirely different type of situation to the amp than a speaker that is just a few ohms higher in impedance ( all other things being equal ). How it deals with the increased need for current and if it has enough voltage capacity so as not to compress or clip under dynamic conditions would become a very system and listening style specific situation.
Keep in mind that Damping factor is the difference in impedance of the amplifer output section itself and the load that it is presented with from the speaker / speaker cable. The lower the output impedance of the amp and the higher the load impedance of the speaker and cable, the higher the damping factor. A lower speaker impedance and / or a higher output impedance on the amp instantly lowers your damping factor. While some say that damping factor doesn't mean squat, theoretically speaking, a lower impedance speaker will always be less "controlled" than a higher impedance speaker. That is why it takes "SHEER MUSCLE" to handle low impedance, high reactance loads. There is no getting around the need for both high voltage and high current levels in such a situation. Since this requires BIG power supplies and a lot of ouptut devices, price goes up accordingly. As such, a "good" system based on low impedance speakers will effectively cost more than a good system based on higher impedance speakers.
Sorry for the rambling, but i had so many ideas that i wanted to convey, i got sidetracked. I hope that you can understand at least part of what was roaming through my head. If i've really put my foot in my mouth on this one, i would appreciate it if someone ( Hi "Bear" !!! ) would take the time to please clarify and correct my mistakes. Sean