Like you, I checked the function of the input attenuator and the 120MHz reference. Both were OK in my case. The tracking generator and attenuator was working along with the associated electronic attenuator. But the primary local oscillator could not lock. It was strange, I got a sweep a sweep but the sensitivity / range was out by c.55dBm, i.e. I had to insert -55dBm for the trace to read 0dBm. The sensitivity changed with frequency and below certain span (10MHz) the signal disappeared. Internal 714Mhz and 21.4MHz signals were present.
See pictures of the error log and sweep with the noise floor averaged and sitting about 55dB above where it should be.
My existing spectrum analysers won't measure above 3GHz and without other equipment I could not check the internal 3.5Ghz signals.
Apart from the errors with the LO loops not locking etc, the LCD display was dark. I thought it was the cold cathode tubes and I looked at these, they were in fact very bright. I suspected that the actual internal liquid had gone dark due to age and that the unit was kept in a hot environment. I used an external monitor as the display, this worked well!
After some time checking I decided to return it to the owner. The original fault had been reported to me as a damaged N-Type input connector but clearly the unit had many more problems. It may well be available on EBay for sale shortly for spares.