Follow the links to authoritative documents within those threads.
Also, what do the 'slippery eels' at National Instruments say? I'd be asking them. Don't get me wrong, they have some super engineering staff, but sometimes it's hard to get past the frontline "is your computer plugged in" guys and girls.
This discussion comes up all too often. Those two-screw RS485 adapters are the reason. Find a 3-screw type, they're out there. The third screw is for your ground.
There are some VERY limited use cases for not connecting the grounds of the systems involved, but my approach would be to only investigate not connecting the grounds if your communications issues exist with ground in place, OR if your industrial environment precludes connecting the grounds for some reason.
As for your fibre experiment, it's going to depend on how the fibre 'isolator' handles the bus turnaround, and whether both the NI CRIO device and your Arduino code are working in concert with the isolator. It's not guaranteed.
Finally,
I also tried to use a RS485 to optical fiber converter instead to use the wires ([CMM/1S - RS485-synthetic fiber converter - Low voltage, but I still get the same problem: it works, but it is very sensitive to noise.
In this case, did you establish a ground between your Arduino power and the fiber converter's ground(that "int gnd" screw)?