This receiver has a tube based PLL for the LO. This was completely off
on all bands due to the fact that all the cores of the Ls loosened,
they were kept firm with some sort of glue across the thread that
pulverized over the years. The vibrations of the drum did the rest.
After the alignment, it is now rock stable, better than the 8640!
However, I did not consider the IF filters when I did this, I tuned
the LO for *exactly* 3.3MHz IF that then become 300kHz after another
mixing with a fixed 3MHz quartz oscillator. It is not a good idea to
tweak this oscillator too, as it also generates the harmonics used by
the PLL. This is actually 3 to 5 Hertz off.

I am *very* worried of upsetting the filters. I like the idea of
identifying which trimmer changes which part of the curve before
attempting the real job and preparing a reference chart. I am
convinced that some previous owner tried the alignment before and not
necessarily did a terrible job. The manual just has charts with the
expected shapes of all six, the number of peaks, at which freq they
should be, the maximum ripple. The representation on the scope is
linear, I suspect it will not be accurate enough. Today I tried with
the digital scope, set the persistence to 10sec so that I can see a
curve produced by ten sweeps, this seems to be an improvement. Also,
-3dB=0.707x linear, the digital scope has cursors that also help.

Another thing I wanted to ask and you already mentioned: which filter
to start with? The narrowest, this one is easier to center at exactly
the right freq. You make a difference between "crystal" and "tunable",
however this thing must be quite odd, the crystal filter is also
tunable, has its own ten trimmers as the non-crystal types.

Thank you again for your help
Best regards
Paolo