Motor/signal ground isolation

I realize that this is a topic that has been addressed in some capacity in a few other threads, but I'd like to just clarify it further. I'm specifically using the SBL23xx series, but I imagine this question applies to several different controllers from what I can tell. It's regarding how it's advised in the manual to not have any external connection between IO connector GND and motor GND.

Well, I understand that one concern is regarding loss of the (thick gauge) motor ground connection on the tabs, making the return path for the motor current go through the IO connector, potentially fusing wire and/or destroying the controller. I agree that this is a legitimate concern, but it seems that it could be accounted for by having two separate thick wires connecting the battery negative bus to the controller tabs; that way both would have to come disconnected for that to be an issue.

The other concern, though, seems to be regarding noise. I know it's a complicated subject, but if the whole system is referenced to a low-impedance common ground, taking care to avoid ground loops - have you found in your experience that it's still better to isolate grounds? I guess what I'm confused about is that it seems the IO connector GND is connected to the motor GND anyway (correct me if I'm wrong - I'm basing that on measuring low resistance, i.e mohms, between them with multimeter) so I don't understand where the noise benefit is coming from by isolating?