Thanks guys, some interesting info and ideas. I'd particularly like to explore the flame sensor that detects UV. It would be great if anyone has more info on this.
After doing a bit more research, I think that's a good idea.
According to Wikipedia, a fluorescent lamp can fail, due to loss of mercury, which can happen when all of it is absorbed into the glass tube. If this happened in a germicidal lamp, it would still continue to emit visible radiation, but would no longer emit much UVC, which would fool a visible light sensor.
https://en.wikipedia.org/wiki/Fluorescent_lamp#Loss_of_mercuryHow about using an ordinary, visible light detector, behind a piece of glass, with the phosphor, from an old fluorescent tube? If the sensing threshold is set high enough, then it won't be triggered by the dim light generated from a tube suffering from mercury loss.
Another option could be both a timer and a visible light sensor? The tube could be routinely replaced,. after 8000h has elapsed or it no longer emits and visible light. Mercury depletion won't happen after just 8000h. It takes much longer than that and isn't common because the filaments normally fail first.