I'll look more into using the tpic6 chips. My plan was to use the shifts to drive an NPN transistor(2N3904) that would actually switch the solenoids. I was going to use a series of shift registers to use 1 pin per transistor solenoid pair. I hadn't yet learned of the multiplexing concept.
I still think this is the way to go. Multiplexing LEDs using persistence of vision is good. The LED doesn't mind being switched on/off many times per second to create the illusion of many on at the same time. I don't think a solenoid would behave well. It wouldn't ever really be completely open or closed during operation. The waterfall display would have many valves open and closed.
The benefit of multiplexing to save on wiring interfering with the illusion in an LED cube is good. Not really a concern in my application. The only thing on display and visible would be hidden anyway.