Perhaps one could easily time how long a given number of light units takes with a brand new bulb, and then time it monthly to start tracking its degradation.
Then it may be a matter of how much longer it's taking being merely an inconvenience, or getting some real facts about whether the UV output diminishes at a different proportion, and eventually even the same amount of light units, no matter how much longer they take, isn't cutting it any more.
Of course, in a few years, LED's (or something newer and cooler) may render this moot.