I rushed to plug it in when I got home and was pleased to see the power LED light, then less pleased to see it fade away as the 100mA mains fuse opened.
I put a 500mA fuse in, which kept the power supply running long enough to measure the unregulated input to its 15V regulators - both were below 10V. I then tried it from a bench supply with similar results, so it wasn't the bridge rectifier or smoothing/reservoir capacitors. You already know what's coming, it's a RACAL from 1978, it can only be tantalum capacitors.
Sure enough, after a few minutes running from my bench supply, where I'd nudged the current limit to 500mA and was watching both voltages falling alarmingly the smell of smoke became apparent despite it being confined to the heavy cast screening enclosures for the two major PCBs.
RACAL had used 16V tantalums on the positive and negative 15V rails. The first to go was a 10µF on the sweeper board, followed by less dramatic failures of 47µF. All red ones, was that STC? I think it might've been. I changed all the more likely to fail capacitors on the sweeper and the millivoltmeter began to show signs of life - the various oscillators began oscillating.
The remaining problem was that it was lacking in sensitivity, only 10dBm on the -50dB range got the meter swinging, additionally I couldn't follow the calibration steps because the signal was too small and seemed to increase only when I slammed the probe offset off centre. I solved this problem accidentally by spotting that my scope's input resistance on the probe output was enough to improve the situation - so clearly a DC offset. After dismantling the probe to check the diodes because I was convinced that all DC offsets much come from damaged diodes I proved myself wrong, the diodes were fine. The offset was yet more tantalum capacitors leaking horribly, so I changed three more on the amplifier/signal processing board and it started working normally with 0dB being within about 0.2dB all the way from -50 to +13dBm (as high as I can go without summoning more RF amplification).
This unit is special in a few ways, firstly it has an olive drab paint job and second, it doesn't match the service manual, because it lacks a 4051 multiplexer and a pile of variable resistors that would allow each range to be fine tuned individually. I expect this was cheaper than using precision resistors and my unit is probably *better* because of this omission, but I don't know for sure. Cost cutting by adding more, cheaper parts?