Haha, I was thinking that the spectrum analyzer would be useful to cal the sig gen, but seems like you have a chicken and egg situation.
Its always interesting to see how HP and others combined electrical and mechanical engineering, especially in older instruments. My Fluke AC calibrator has a front panel with belts and cogs and so on, all working together like clockwork, and of course there is the famous HP 8640 sig gen (I saw one languishing in a local surplus dealers office, on the random kit shelf).
Regarding your missing plastic parts, from the look of them I'd guess that they are made of injection molded POM (delrin). Sourcing salvaged parts maybe the best option, but if I was forced to tackle their replacement myself, I'd be concerned about using a MSLA resin 3D print as I find that it has poor long term stability (though that could be just be due to my lack of skill and using cheap resin). I'd rather try using an FDM (filament) printer, but I'm not sure if it could manage the small details. The advantage is that, although rare and difficult to print with, it is possible to get POM printer filament. Nylon of any sort may not be suitable for this application (due to being too flexible and also absorbing water may affect the attenuator performance), but it's a relatively popular 3D printing material so probably worth a try. A custom machined POM part would be great, though likely rather costly. It would also be difficult to exactly match the injection molded part as molded parts are "negatives" based on a machined tool i.e. it's difficult (expensive) to get a sharp radiused inside corner on a milled part but easy to get a sharp external corner. The opposite is true for a molded part.