Regarding cable shielding - it depends!
I’ve rubbed up against this topic in my day job. However I’m not a Ph.D EE who specializes in electromagnetic compatibility. But, I’ll share what I’ve read on the topic:
The argument against grounding at both ends is that this can be a path for a ground loop to flow. If this occurs, it actually is a noise source for the signal - the current flow can then inductively couple noise on to the signal wires. So, you ground at one end Usually this ground is at the source end, since it’s the reference for the signal voltages. If the signal source is not grounded, then it is better to ground at the end of the load.
The argument against grounding at one end is that the shield would be an effective antenna for noise in frequencies where the shield length is 1/4 the wavelength of the noise, or higher frequencies. Some guidance is that the noise will start to couple after 1/20 of a wavelength.
The other argument against grounding one end is that high frequency noise will be “conducted” across the stray capacitance of the cable anyway, and you’ll get current flow across the shield to the connected ground even though it’s not well connected at one end.
The best scheme is to do a combination of both. Have twisted pair signal wires with a double shielded cable. The inner shield is connected to ground to mitigate low frequency noise, then that is shielded by a high frequency shield connected at both ends.
In the drone application, I guess the question is, what noise are you concerned about, and where is it entering the system? If it’s just between the different electronic components, I would just try an unconnected shield.
A drone is battery-powered, so everything is electrically floating, you won’t get ground current loops. The biggest source of noise on a drone would be the high power PWM signals feeding the motors from the ESCs. I would try just shielding those wire segments.
I would think you could continue that approach using small metallic enclosures for other components, but I’d be careful - you don’t want to short anything out.
In fact, unless you have evidence of a noise problem and a way to measure it, I don’t think I would “make it better”. If you do have a problem, then I’d use an oscilloscope to measure the problem before and after each improvement. With a systematic approach, you’ll land on the best solution in your application.