Well OK, but it's still a work in progress so not everything might make sense, and not all the values are final. You can see why I wanted to get you to imagine a standard type of supply....
I've added the proposed diode/resistor to the channel at the top (B) but not the bottom (A) so you can see both versions.
The rectified & smoothed output from the mains transformer can be up to 55V, (50v nominal, can go as low as 45v) the pre-regulator should be outputting between 3 and 5v more than the final output (it was 5 but I intend to re-calculate to get 3).
I chose to use a sziklai pair for the output as an experiment, and I've stuck with it because I like it - it's effectively an NPN emitter follower with almost as much gain as a darlington. It's the small driver transistor that suffers the vulnerability in question, and what happens as far as I can work out is that it can fail so all pins are shorted due to it's Vbe being exceeded, which then gives a path for excessive base current to flow in the PNP (or exceeds it's Vbe because the input voltage has been lowered, take your pick), completely shorting it also, and also destroys the control transistor into the bargain (which fails open collector) because it is exposed to too much current. I'm running out of suitably rated old transistors, I don't want to start blowing up new ones!
When the voltage is turned down, the darlington control transistor conducts bringing down the output. However if there's no load, or a very light load, it will try to pull the voltage all the way down to zero until the output catches up. Now suppose the output has been turned up to 40v, or the high 30's at least, and then I decide to turn it down quickly. So now with the diode arrangement suggested by ron, there is a capacitor with 40v sitting on it, discharging through a diode and through the fully on control transistor. Obviously quite a lot of current can flow there, which will definitely destroy the transistor, hence the resistor is there to limit it. If you look at the datasheet for the BCX38C, and use the handy calculator
here, you will see that with 220 ohms this keeps the transistor just barely inside its safe operating area as the voltage falls.
I put the diode back across the output transistor into the schematic ages ago, but like a dummy I haven't implemented it yet.
The strange looking 2 zener (because they are the values I happen to have, no other reason) and resistor combo is there in case the control transistor fails O/C, it will keep the voltage from the constant current source down to a little over 40. The tracking design would work just as well with a resistor providing the base current, but the CC source gives some ripple rejection.