Correct, at first, the pilotone was used to control the SEPMAG machine, later the switching was reversed so that the tape speed was controlled after the signal because it is easier to monitor the tape recorder and not the magnetic film drive with its heavy flywheels.
I think you can go far with crystal control. Standard says maximum deviation of ± half a frame per 16,000 frames. Crystal control goes well with digital equipment, the CD works on a crystal-generated and downcounted signal. If you see a way of proofing the crystal control, accuracy might be still better like ± a quarter frame per 32,000 frames. I cannot tell you the exact way of doing it but a computer’s crystal frequency can be captured (before or after downcount, I don’t know the details) for a common reference.
Perfect synch is of course attained when both synchronous motors of picture and sound camera run on the same mains outlet.