Limits and initialization in a restricted area ..

Long ago in a galaxy far far away I used a similarly sized /constructed robot arm as a tool changer:

This project was done with the origional verison of the Dorna robotics dorna 1 arm. The primary reason I'm no longer using it is: The robot is in a restricted area and getting the robot to zero the axis(s) is problematic. It also had issues with reliabilioty being open and getting swarf from the mill. I'm thinking about redoing this project with an AR4. Looking at your docs and videos I think I'd have exactly the same initialization issues with the AR4. It strikes me that one soution would be to add a 360 degree I2C or SPI based magnetic position encoder on each axis. This is not remotely accurate enough for the formal zero, but its accurate enough to grossly preposition the robot arm in to a safe orientation where one could pick the order and position of joints to sequentially zero the individual axis in a way that would work in a positionally restricted environment. Probably grossly point the arm straight up, zero the azmuth J1 then zero joints from the end effector back from J6 to J2

Another possible solution is to make a 3D printed optical interuptor on each axis where "Zero" is the change from blocked to unblocked. That way you always know the correct direction to drive an uninitalized robot toward the "zero"

IE if the interruptor is blocked drive the joint in the - direction if unblocked drive in the + direction... etc... Has anyone done any experimenting or work on more robust robot arm initaliztion in a restricted area? I'm also interested in thoughts on the robustness and reliability of the AR4, and ability to make it swarf proof... IE the pictures show exposed limit switches and belts that I would probably want to print covers for, has anyone already done this? In all looks like a cool project.