If you are running your own solar array, though, you will vary your grid load on much shorter time scales
I'm not so sure that these latter variations are significantly less predictable than what you mention in the former paragraph. Isn't the output of a solar array dependent on the insolation? Isn't the insolation dependent on the cloud cover? Can't you predict the cloud cover in any single place simply by taking advantage of real-time meteo satellite data? A similar feedback could be established for wind power. Given enough data, I'm reasonably certain that models could be established that would allow you to predict how the solar and wind power generation distribution is going to change in the next hour(s) so that you could prepare for it.