My thought is that a device's loading on the bus is expected at each node, including the terminated ends.
Take a look at this image showing how a CAN bus is properly terminated (credit: https://www.allaboutcircuits.com/industry-articles/overcoming-can-design-challenges-can-signal-termination-made-easy/)
![enter image description here](https://cdn.statically.io/img/i.sstatic.net/2LZae.png)
Just looking at the schematic, what is that far-end termination resistor connected to if you remove the device at the far end?
![enter image description here](https://cdn.statically.io/img/i.sstatic.net/HqfUC.png)
Well, it's still connected across all the other devices. So if you remove that far-end device, the new far-end device with the termination is now, in your application, the next closest device to the termination resistor. And you still have not really strayed from the protocol.
That is the beauty of CAN and other differential protocols like it. The key is to have the termination at the end of the line. It doesn't really matter if the end of the line is 8 feet away from the devices or 8 mm away. You will experience the same loss whether there's another device next to the termination or not.
What you don't want to do is put the termination across one of the first 3 devices and then have another device 8 feet away. Although 8 feet isn't all that long in terms of CAN data rates, that extra 8 feet will be an open stub that creates reflections at high-frequency harmonics and picks up unwanted EMI.