Hi David,
I have to admit that using visible light was probably a stupid idea. There seem to be some more photoresists that are somewhat sensitive to visible blue/violet [1], but they seem hard to get.
So back to the idea of using a chip socket so the DMD can be replaced when UV breaks it: This will allow to learn not only about how well the DMD takes UV, but also about the failure modes. For example, blurring the image a bit is probably not as bad, and even faulty pixels are covered by typical design rules (taking MOSIS SCMOS as an example). OTOH if the electronics of the DMD get faulty, it's probably game over. Those are things the datasheet probably won't tell you, so trying it the only way to find out.
Another idea would be test patterns using visible light. If the DMD degrades gradually, a test pattern would allow to find out if it is still usable or not. By that, a degraded DMD is no longer a catastrophic event, but rather is part of the expected lifetime of the DIY kit and allows to factor the price of the DMD into the price of the service, mapped over the number of dies produced with it. For example, if the DMD can shoot 100 images before it breaks, and costs 50€, it's 0,50€ per image or 2€ per die if you need 4 photolitho'd layers.
(BTW, the datasheet only says that bright UV and simultaneous temperatures > 70°C are harmful, so that might mean that nothing bad will happen if you just don't stress the DMD too much)
Greetings,
Martin