So I checked out the 8 V supply feeding the 5 V supply - it looks really
bad! Similar waveform to the 5 V rail but dips from a high of maybe 14
V down to about 2 V. That's at the junction of the cathodes of
rectifiers CR820 and C821, C821, R821, and the collectors of Q829 and
Q835 on the Rectifier Board.
If either of the rectifiers were shorted, you would see the waveform go negative across C821. The minimum value of 2v indicates they are OK.
Then I checked at the anodes of CR820 and CR821; more convenient to look
across C820. I saw about a 30 Vp-p near-sinewave there, with the anode
of CR820 180 degrees out of phase with the anode of CR821.
This is OK.
I tried to measure C821 (18,000 uF) in-circuit with my cheapo Chinese
transistor/resistor/diode/capacitor checker, but it couldn't get a
reading. Ditto for C820 (0.1 uF, 100 V). I assume the transformer
effectively presents a short and throws off the checker.
This is what I would expect. Don't expect in-circuit measurements to be meaningful with respect to a single component in a network. The measuring instrument can't sort out a single component, as it see's everything connected to it and makes a futile attempt to make sense of the measurement.
Also, R821, from the cap/diode cathodes/transistor collectors point to
GND, reads 5.5k ohms, when it's supposed to be 4.7k +/-10%. I was
surprised to see it read high; most times the sneak paths make resistors
read low in-circuit. Not that 5.5k probably makes a difference.
Either due to residual charge on C821, or R821 has drifted high, which is not unusual if it is an aged carbon composition resistor.
Anyway, should I suspect the cap C821 or one of the diodes CR820 or
CR821? Or something else altogether?
All the info you have provided points to an open C821. Easy to confirm by tacking an additional electrolytic cap across it and looking at the waveform again. Even a small value, say, 100uF should make a measurable difference. If the lowest level, 2v, increases, then you have confirmed it is likely bad.