Assuming the fiber optic switches had their ports in the same position, why not just make some tooling to hold the cables in position relative to eachother and turn 600 cables into essentially one 600 pin cable? Or n 600/n pin cables?
* This doesn't get you out of cleaning the connectors.
* You'd need very precise alignment to make this work, and I'm not sure our manufacturing tolerances were up to it.
* We would have needed 20 of whatever setup we used to do this. If you're talking in terms of robotics, that's probably the big deal breaker.
* Cables can still break, and getting them out of whatever holder you put them in could be an issue, not to mention even finding a break buried in 600 cables.
A lot of it really just comes down to "fiber optic cables are fussy and fragile," really.
Clean the connectors while they are in their fixture. This opens up the possibility of cleaning several simultaneously.
Making something more precise than a human hand is not particularly difficult. If you don't have the manufacturing capability to do it, there are plenty of job shops out there that can for a reasonable price.
If you can make the connections faster, you can run more in series, so you probably don't need 20 sets. Even if you did, making copies of something is cheap compared to the initial development. External hardware can be set up to service multiple stations.
Many cables moving together in properly choreographed ways are going to be far more robust than an intern fumbling around. If you're making the tooling such that it is difficult to do routine maintenance, then it's a poor tool design.
In general, fussy and fragile things are the best for automation - you're already bending over backwards to deal with it. It's when things are easy for humans, with their amazing ability to adapt to circumstances and immense contextual knowledge, that automation struggles to produce competitive results.
