If you only produce just enough PhDs, then when you start the next Manhattan project you need to scrape the bottom of the barrel and grab someone who's been out

If you only produce just enough PhDs, then when you start the next Manhattan project you need to scrape the bottom of the barrel and grab someone who's been out of the field for 10 years, or wait several years for the rate of PhDs being completed to increase. If you produce a substantial excess, on the other hand, you can instantly scale up to meet increased demand. It's all about that pipeline. But yes, moonshot projects can also be utilized so that past PhDs can be kept sharp. Indeed, a huge number of science programs currently do exactly that - society may see little direct benefit to discovering methane in the space around Uranus, but the people optimizing those radio telescopes are pretty nice to have around if you think you'll ever need people to optimize radar systems.