Thank you for the hint, where the 50 ohms are placed. I measured them and have exactly the same values like you. So that's not the problem.
The error is sporadic and in case the signal is getting through, sometimes only with significantly wrong levels (20-40% less). This could indicate the relay with an increased contact resistance. But I don't have a failed self calibration test. According to the schematic, the self calibration lines start behind the 50 ohm switching relay?
What I find confusing is, that the schematic:
https://w140.com/tekwiki/images/f/f9/H2462_schematic.png does not show all the relays. Specially the black one in the 50 ohm switching path.
I don't see a chance, to measure the contact resistance. The contacts are too hidden and even, if I reach them, it is not sure, if the measurement really gives the right direction. So I problably have to replace the black relay without a 100% evidence.
The form factor of the black TN2 relay is not available anymore from a big distributor. Only option is, to take an ebay offer with new samples from china or original, but used ones.
What is your recommendation?
The yellow relays might not be the problem, since it is independant of the attenuation level and it should throw a self cal failed then.
Anyway, I have to remove them, to remove and solder the new black one.
Would you replace the yellow relays at the same time prophylactically or keep the old ones?