Jos, I've never seen a written source with all needed information cooked down.
As this doc linked below describes, the way of working is, follow guidelines created by eariler mistakes, and hope for the best ;)
The theroy is easy; avoid all parts and connections that can behave nonlinear at high RF power. High contact pressure, silver against silver, no magnetic materials, no oxides and so on.
Some parts are easily understandable, like that red copper oxide has semiconducting properties and was actually used in recifiers. The "rusty bolt effect" is also well known in medium-wave transmitter installations. But there seems to be other, not so good known mechanisms as well, more subtile.

The doc is in english, but the file name is in swedish.
It is published by the Swedish space authority, and the file name is ( translated): characterization of passiv intermodulation in space communication antennas.

https://www.rymdstyrelsen.se/contentassets/de067a79466749efa22b953340e47293/10.-karakterisering-av-passiv-intermodulation-i-rymdkommunikationsantenner.pdf

This slideshow illuminates printed circuit board properties in a PIM context:
https://www.slideshare.net/manfredhuschka/the-influence-of-rf-substrate-materials-on-passive-intermodulation-pim