This spawned an unrelated thought. Would using this source of input (a fish tank) be a good source of entropy? Would seen to me tracking fish movements would be quite random.
I dunno, I would suspect that fish would exhibit patterns when averaged over a long enough period. They probably patrol the perimeter regularly, check the same spots for food every few hours, some are territorial and will patrol their spot. They're probably not all that random. Living things tend not to be.
Repeating patterns are presumably observable on a longer time scale. If this is the case, there's multiple ways to separate higher frequencies which should provide higher entropy.
Funny - after watching this video a friend had this same thought about the lava lamps and random numbers, then I had the same thought about using a fish tank to generate random numbers.