I managed to repair it. The fault was a shorted diode near the output. I suspect it's a protection diode (looks like a 1N4148) that normally clamps the output signal so it does not exceed the supply voltage.

I also replaced an electrolytic capacitor, probably a smoothing cap on the negative supply line.

I might publish an update after I run the calibration procedure (although the manual describes a procedure using a spectrum analyser, so it might take me a while to find one somewhere).