The data sheets all spec stability when the output has a fraction of one uF on the output, but for a variety of (bad) reasons some designers insist on putting, sometimes, tens thousands of often very low ESR uF at the output. That's extremely hard on the protection diode and it'll have to be beefy, I'm sure at some size of cap it'll be bigger than the regulator die, and none the less someone will attach an even bigger cap (why?) and blow out the reg anyway. In the old days regulator ICs were very expensive and caps were cheap, so you'd try to run entire systems off one expensive reg and one HUGE cap. Now, reg ICs are very cheap and small and caps and especially heatsinks are expensive and relatively large, so its cheaper to regulate multiply and locally, with small caps, no need for the diode.. so don't expect changes from the mfgrs any time soon..
Assuming money were no object and people would pay twice as much for a reg where the die is half protection diode, the next problem is a reverse biased diode is a RF short, so goodbye 80 dB CMRR or whatever it is, and uV level output noise performance. The old 78xx series is quiet enough for analog or digital use so they'd need to sell quiet ones for RF and protected ones for digital. Tens of mV of RF noise at 5 MHz is fine for digital, not so good for a microphone preamp. I'd guess that "most" IC regs in a circuit with a giant cap are just driving a giant transistor anyway to get 15 amps out or whatever, and that junction is the one needing protection anyway. So the lm317 in my giant astron 12V power supply will never blow... its "only" driving some giant pass transistors, which are the junctions actually needing protection diodes. So most of the time circuits with giant caps come with giant transistors making protection diodes on the die a waste. The 78xx series is short circuit protected and they can sell that with a straight face without a reverse protection diode. With a diode, all you need is the drop from input to output to exceed its PIV (power spike? Shorted output?) and it'll zener till the cows come home, blowing everything on the board with no short circuit protection until the protection diode melts. Now the hearsay handwavy starts with "due to design of the reg, the overall system is more survivable without the protection diode". A low PIV might make the reg useless for HV applications. The 7805 for example doesn't actually output 5V it forces the ground pin to 5 volts lower than the output. If ground is actually 995 volts, the output will be a nice regulated 1000 volts for a geiger counter or whatever. That means the reverse protection diode "needs" a PIV like 1000+ volts. On die, thats tricky and expensive. Its best left off die for the engineer to optimize. Finally in my experience its mostly theoretical anyway. You can't have reverse current without a complete path, and the reverse path tends to be pretty low current when the power is off.
