There might be one reason they let this spin up and down.
We mostly power the grid through a bunch of constantly humming systems, but our usage has all kinds of peaks (diurnal, seasonal... the British grid has a massive peak right around the time Eastenders ends and everyone puts on a kettle[1]).
If you are adding some power to the baseline, that's helpful. If you're promising the grid you can add some power during one of the daily or seasonal peaks, that's almost as valuable as adding a new source of always-on power. By providing energy for even just six hours a day, you might save the grid from requiring a new always-on plant. So in terms of opportunity costs, even by being twice as expensive moment to moment, you're still cutting costs in half (unless everyone followed your strategy).
So there's probably some number of hours of shutdown where it's worth the coal, some number where it's not, all contingent on what plants are nearby, and how close your customers typically are.
Related but tangential, one of the unintended benefits of electric cars would be adding a bunch of consumption at night, smoothing the diurnal cycle, letting us pursue more lower-cost always-on generation. (This is before you even consider distributed storage proposals that let the grid draw from parked cars in emergencies.)
Predictability is almost as good as a natural resource.
[1] http://www.bbc.co.uk/britainfromabove/stories/people/teatime...
In parts of AZ there are serious brown outs and rolling blackouts in the summers as A/C kicks in everywhere, which is probably the biggest draw.
I'm opposed to coal altogether for pollution reasons alone. It's bad enough between the cars and airport traffic, don't need more.
Interestingly, the only article on a Phoenix area blackout due to high load I was able to source happened in the winter... probably caught SRP totally off guard:
http://www.azfamily.com/news/local/Rolling-blackouts-sparked...
Note that's when running at "maximum power". You don't have to extract the heat that fast. Given a hypothetical and utterly irresponsible grid supplied solely by this design, you'd overbuild to the point that you'd have enough stored heat to last all night.
This is a matter of tradition and convenience. It is quite possible to use a just-in-time steam generator.
In fact many steam power systems are vapor phase only (non-condensing) to avoid corrosion and droplet impact damage. Steam is used because it is cheap and nontoxic.
Nuclear reactors are easy to turn on and off once a day. Naval reactors do it as a matter of course. There's just no point. If you build a grid scale reactor, you might as well leave it running all the time, at which point you optimize away the rapid start capability.
Any power plant that uses steam turbines is going to have a long ramp-up time to boil a huge quantity of water needed to produce the steam. Nuclear plants do this, and I imagine most/all coal plants do as well. Hydroelectric plants don't need to because the falling water drives the turbines directly.
The first thing I thought of when I read the article is that this power plant can't be very efficient. It's using steam turbines, but six hours after sunset there isn't enough heat to keep the turbines going. That means they need to reheat every morning. It seems like it would have been better for them to include a coal plant for nighttime operations to keep the steam boilers hot and keep the turbines going all night. That would be flexible, because the coal burners could start up quickly to keep the already-hot water boiling. (Can't use nuclear this way because you can't really turn a nuclear reactor on and off.)
I'm totally guessing here, but I suspect they didn't include an alternate nighttime operations fuel because this is a "100%-Green" project that either wouldn't have gotten funding if there was a coal plant involved or it was just intended as a demonstration of the solar capability and didn't need to be practically efficient.