Nigel,
I do recall that issue we faced. It was the reason why we used a pedal check during run-up to look for the problem until we could get all the bad servos out of the field. It was not caused by turning off the hydraulics, it was uncovered that way.
The problem was that one of the tail servo sections had seals that wore very quickly (I can't recall if this was a material or an assembly problem) and allowed too much blow-by, a condition like badly worn piston rings. In this condition, a single worn servo would not develop enough force by itself to control the tail rotor. Since we have a dual tandem servo, and since either piston can fully control the tail rotor, the condition was completely masked in normal dual operations.
That problem was found and fixed a long time ago, but I was on the other end of the S-76A system, working to help identify it, find a way to locate the problem, and of course, fix it. Small world, Nigel! It was all fixed with a change to the tail servo seals, and has not been an issue since.
When you turn off one servo, you get the other as the sole force provider. If that servo was worn, you can get the pedals to drive because the tail rotor is going to a pitch setting it likes, and the only thing stopping it (the good half of the tail servo) was shut off.
It actually was not a servo hardover, it was a lazy servo, but that let the tail rotor walk back to the neutral position (sort of the zero thrust position).
Why does the tail rotor need a servo? Because the blades are all producing a pitching moment as they are moved to new collective pitch angles. At high thrust, the pitching moment is quite appreciable, and the servo is your way to muscle the blades to the position you want.
Why did the collective drive? Recall that the collective and yaw are interconnected, and if the force is very high on the pedals, and you push on them with a bunch of leg force, that push will be reflected off the mixer and come out on the collective. Yes, it was you who helped drive the collective pitch, because the yaw was frozen, and your massive leg force had nowhere else to go. There are corners where the collective will drive by itself, but the collective will have lots of range in the middle flight position.
Did the cyclic freeze? No, but when the collective drove and the pedals leaped, who can blame a pilot for not doing a staff study while he leaps to move the switch back?
Why was it not seen when the rotor is not turning? because it is not a hardover, jus a lazy servo, so it needed some appreciable tail rotor force to show itself.