There was no obvious leakage to have caused the corrosion.
Sadly resoldering did not fix the problem.
I gave up and bought me a Siglent SSA tracking analyzer.
It is not as good as the HP, the phase noise is 10-15 dB higher.
It shows especially at frequencies below 500 KHz, where the noise level increases a lot, because it is mostly showing its own phase noise.
HP must have used the LO offset and FFT in a better way, because it could measure close to zero Hz without its own LO noise disturbing the measurement.
It could also be that the Siglent first mixer is out of balance, letting noise through.
The HP was also better at showing peaks at low bandwidth combined with a large span.
At least the Siglent is better/Kg
