

Out of interest I did some estimates and it seems that an asymmetry of three billionth of the total thermal radiation would be enough to rotate the probe once over a timescale of 10 years. So if the radioisotope generator has even just a tiny bit of a different infrared brightness on one side, it would turn voyager in a few years.
notes on calculation
Voyager weight: 815 kg
Approximate Diameter: 1 m
Assume mass and thermal radiation emitted with a center distance of this diameter. Then we can calculate as it would need to move 2π 2 m. It should be enough as coarse estimate and underestimate the acceleration. Distance to move: d = 6.3 m
Assume constant acceleration due to thermal radiation
RTG power at start: 3 * 2.4 kW = 7.2kW
RTG power now: 7.2kW * 10^(48/88) = 4.9 kW
Total of thermal radiation: 4.9kW / c = 16 uN
distance moved: d = a t^2 / 2
assuming 10 years accelerated movement movement:
a = 63 mm/yr^2
F = 52 fN
3 * 10^-9 of thermal force
The AI needs help to cut the loop, perhaps it needs a new set of knives?