My take away we have very short buses in our robots by CAN norms and can get away with behavior that would be bad news in most cases. However you are on your own, a scope may not be enough to catch glitches that may cause problems. Even if there are problems higher layer error correction may catch it. None of that means your robot will or won’t have problems on the field.
Pete makes good video’s I recommend anyone interest in CAN or ODB look at his videos.
Our robot had a loop of the CAN Bus run up the elevator and over the arm. During one of the competitions the “return” leg of this loop broke causing some but not all of our CAN devices to fail. We found that if we disconnected all but 7 devices, the remaining ones would work just fine without terminating at the PDB.
Moral of the story: CAN is unpredictable when it’s not wired to spec, and incidentally that makes CAN breaks frustratingly hard to diagnose.
It sounds like you are saying that you routed a twisted pair out to CAN devices on the arm and continued the pair back down the arm and elevator to get back to the PDB. It would be simpler to put a 120 ohm resistor as termination at the end of the primary pair at the end of the arm after the last device and then eliminate the entire return pair. It is unnecessary to loop it back to the PDB. That is just a convenient place to terminate if it is near the end of your loop.