Well things didn't turn out quite as expected. Replacing U304 caused the original fault to re-appear - very low voltage +5A, no clicking relays on self test, and the return of the set of failure codes associated with error 608. I started checking some of the other supply rails, and I found that +15A was at about -0.6V. This is weird, because U304 is shown as powered by +15B, which was OK. But I found the schematic in the service manual was wrong, and U304 is indeed powered from +15A.
Both these rails are derived from the +18V rail by means of 3.3V zener droppers. CR305, which generates +15A, had failed open circuit. Bridging across it with five 1N4148's in series (since I didn't have a zener handy) caused the +15A supply to rise to about 3V, and the diodes to get very hot. Since so few components are connected to this rail, I relied on inspired guesswork and unsoldered C322 (basically because it was the easiest place to start). This is marked as a 22uF 25V tantalum capacitor - but it measured 29 ohms! I temporarily replaced it with a (through hole) 22uF aluminium electrolytic, and using the diodes bodge I now get 15.3V on +15A. The self test still fails, but this time only on test 620, which is AC RMS zero. Since there are various wires soldered to different parts of the board, and it is not mounted on its screening plate, I'm prepared to assume it's just picking up mains hum from somewhere. I hope.
I suspect that the original failure was C322, which then caused CR305 to fail, and then eventually U304 died. I think my replacement U304 is still OK, as it only saw the bad supply rail for a couple of minutes. I've ordered replacements for C322 & CR305: when they arrive I'll put it all back together, and if it passes the self tests and performance check I'll soak test it for a few days. I have another DG411 in case U304 has actually been damaged.