quote PRR: "I was reminded today that the VFB (or CFB) op-amp is NOT the ultimate amplifier for all uses. We thought it was heaven back in 1972, but this was folly. Op-amp theory requires that we assume infinite gain: this never happens. It may happen close-enough for many uses, but not really good enough for others. And many times the only reason to dip into op-amp theory is to avoid actually thinking what you are doing. There are many other amplifier building-blocks; unfortunately, dollar op-amp chips drove most of them out of our awareness.
Figure out what you really want to do. See if you can do it in 3 or 4 transistors."
Amen. Op amps helped many people to believe they were analog engineers, and made well-deserved money for companies like Philbrick and their successors like Analog Devices, Fairchild, National, et al. The half- or smaller-fractional truths that made the first-order theory so seductive fostered a sense of power and helped a youthful generation pooh-pooh the preceding couple, who had learned to do more with less, especially important when tubes were the only option and still of some value when transistors were expensive.
Unity-gain-stable voltage-feedback op amps are wonderful for quick and dirty jobs, and may well satisfy all of the required performance criteria. They can always be made to oscillate with sufficiently wacky feedback components, but require significant provocation.
If one has access to the whole circuit including the interstices of the forward gain path, the additional degrees of freedom can be used to advantage. Since this access is usually only granted nowadays via our beloved discrete designs, that is one reason for our tendency to think of these as superior. But there are many advantages to monolithic fabrication, especially those of inherent parts matching and thermal tracking from sheer proximity. So the best of both worlds is a bit of both.
A sad aspect to much of this: for some really interesting things where one attempts to combine high frequencies with high accuracy at low, you simply can't get the d*mn parts---you have to have the formidable resources to pay for monolithic design and fabrication using the latest and greatest "processes."
Fortunately most of audio design, at least analog audio design, doesn't suffer much from this restriction.