[WaveyDipole]

Check the diodes in the rectifier/doubler. The designation EM513 on the diagram suggests a module of some sort and if its potted then a repair might be somewhat more difficult. However, it should still be possible to measure across D301/D302 and D304/D305 pairs. The results should be the same for both pairs. Two diodes in series should test the same as a single diode, just that the resistance or voltage drop across two would be double that of a single. If the two readings vary very significantly (say by about 50%) then one of the diodes in the side with the lower reading is s/c.

I spent a little while this afternoon drawing the circuit in the LTSPICE simulator - simply because I had nothing better to do and hadn't played with it for a while - to see whether I could replicate the problem. For example, the simulator showed 6.8v at pin 4 of the IC, which is close to what you measured. On the other hand I couldn't match the voltages at the base of T301.

Since you seem to have ruled out just about everything else, I tried simulating a shorted diode in the rectifier/doubler. It didn't seem to matter which one I chose, but so long as I only shorted one, the simulator produced a result that might explain what you are currently getting, although there is also a horrible amount of ripple.

I have uploaded a couple of screenshots. The first one shows the PSU working as expected. The yellow trace is the output, while the blue one is the voltage at the collector of T301. The simulator shows around 190v-200v at the collector of T301 so a bit higher than the 163v on the diagram, but the output is -1kV as expected. The difference might possibly stem from the characteristics of the transistor and diodes used for the simulation. Playing with the values of the two parts of R318 to reflect different wiper positions does vary the output voltage as expected.

The second screenshot shows what happens when one diode is shorted. I'm not quite sure how a DMM set to DC volts would behave under those circumstances but those oscillating voltages do seem to average out into the right ballpark.

NOTE: The reason that the amplitude of the AC wave generator is set to 680v is because the simulator expects the peak voltage rather than the 481v RMS voltage shown on the diagram. To derive the peak voltage, the RMS voltage is multiplied by 1.414.