Output transistors run in parallel need to be trimmed to work properly, and they are now up to trying lasers to get this right (Edge), which they really can't. If output tubes are used, they have to be properly biased (or self-biasing), and they are often not at peak bias in practice, thus introducing hum and other discontinuities into the sound.
The more output devices you have, the more collective residual noise you have. This becomes an issue with a system that features highly revealing upstream components, top cabling and top speakers, all fed by a well designed A/C power supply and set up carefully in a good room. Extra output devices are simply not needed if you have the ability to buy sensitive speakers with a benign load impedence. In addition, the more output devices there are, the more one is likely to fail (output tubes are generally easy to replace, but output transistors in well-known SS amps as young as ten years old have been known to go out of production, making the amp a door stop).
I qualified the Singer anecdote thoroughly -- read what I wrote again.
While a monster amp is desirable with inefficient speakers, this is just not the way to go in my opinion -- partnering sensitive, easy to drive speakers with an amp that uses a simple circuit featuring a minimum number of output devices and the very highest quality parts will yield superior sound. High-powered amps have a use -- digital home theater systems, where maximum dinosaur stomping seems to be the goal -- they are a poor choice, in my opinion, in high-end, two-channel analog-based systems. In short, I stand by what I wrote.
PS - My amps run 300B's -- now that's a linear amplifying device.
