I am planning on using SMA connectors for RF with 90 degree cables for most connections because I was able to get lots of them cheap. I'll make up other cables as needed.
I plan to use boxes made of galvanized steel. It's easy to solder, has fairly good H and E field shielding properties and i can get all the scraps I want from an HVAC shop. I'll probably build some small bending brakes of a piece of steel channel and angle for a few standard sizes so that all I need to do to make a box is cut the sheet, clamp it in place, fold it and solder the corners The board will mount to the lid using standoffs with connectors, controls , etc on the underside of the the circuit board. Each size will require two brakes, one slightly larger than the other so the two nest properly
Once a module is finished it will be soldered shut with connector labels on the top and a schematic on the bottom.
Power and control is still an open question. I want feed thru capacitors on power and signal lines but have not identified a suitable item. They are difficult to find at affordable prices. Generally power would be 12 V but other levels as needed. I'm only interested in very low power (QRPp) operation. In some cases e.g. an MCU control block, there will be an internal 3.3 V regulator in a subcompartment.
MiniCircuits sells connectorized version of their parts which package the device in a metal enclosure with the connector of your choice. But at eye watering prices.
U.FL is only good for RF test points as they are only last for a very small number of connections. I plan to place them in strategic locations that can be jumpered in w/ zero ohm resistors.
So that's the *theory*. It will be interesting to see what reality is like. I'm sure it will evolve over time. I got the general idea when I opened up my HP 8601A sweeper. It's a standard set of die cast boxes which are sized to fit a standard chassis. The connectors are on the lids and the boards are edge mounted to the lids.
With regard to measuring insertion and return loss for the connectors you're using I *think* that you could do that with an impedance bridge and an appropriate test jig. However, I really don't think there is much reason to even consider connector insertion and return losses at these frequencies. Surface oxidation resistive effects probably dominate. In any case, to be useful you'd have to be maintaining a constant impedance level everywhere. It looks a lot like a rabbit hole that would consume a lot of time. Before you go too far in that direction, set up the algebra and look at the magnitude of the effects at audio frequencies. If the connector losses matter you'll have to calculate the impedances all along the signal path.
I just bought an HP 8753B VNA which covers 300 KHz to 3 GHz so I shall be going to VNA school for a while. I've also developed a bit of an OCD issue with EMI, so it is likely to get augmented with another instrument that covers lower frequencies so I can properly characterize EMI filters. That gets *very* messy because power system impedances are not well defined and definitely not 50 ohms.