I don't think one needs shielding for the analog one. It has low impedance as far as I can see. If I would see issues (yes I have still to test this config with a longer cable) one could average the measured samples over several cycles to avoid unwanted switching off caused by some spikes. But the tests I did (long ago?) proved that it is quite stable on a low voltage and that it shows a steep increase if there is something detected.
Of course one could use the potmeter and the digital signal, but then you have to calibrate them. I've tried to find a good trigger value. In fact it hasn't to be that accurate since it's no measuring device.
The NTC I wanted to be on a separate (twisted) pair because that one has a higher impedance and could be infected with spikes if next to a stepper cable for instance. But again this could be flattened out over several samples.
Perhaps I will do that right away for both ...