It works if your load is a fixed resistance, the issue comes when you have inductive wires leading from its output, and capacitances in what your powering,
This is what a previous poster hinted at with frequecy compensation, Its not uncommon for larger PSU's to have a small network of passives to make sure "the gain falls below 1 before the phase shift reaches 180 degrees
Essentially your op amp and darlinton have some propegation delays, and response slew rates that mean the device does not react immidiatly to a change in the measured output voltage (e.g. your powering a flashing led, every time it changes state, the amount of current drawn changes, and the control loop needs to react)
At some point a change in the measured input is so fast that by the time the PSU begins reacting, the signal has already started moving in the other direction (e.g. the inductance of the wiring ringing), if the output has any amplification above the frequency of this change, it will build and build until it oscillates,
So this is why you will find filters or similar on the sense point, e.g. a 1nF capacitor, after the divider, this reduces the amplitude of higher fequency signals so that the power supply behaves to ringing or similar, at the trade off of slower responses, (There is a lot of math if you want to fully understand optimising this),
There are other quirks aswell like limiting the op-amp slew rate by fitting a capacitor between the output and the negative input, e.g. you have a big beefy power darlington, Its incapable of reacting as fast as the op amp input, so you might fit that capacitance, to slow how fast the op amp changes its output to be closer to what the transistor is capable of, again to reduce the risk of oscillations, or ringing on step changes.
There are other methods aswell, but those are starting points, confining your device to 1. not try and react faster than your power element can, and 2. behave under "reactive" loads and under any frequency of step response.