This has repeatedly resulted in the ISO finding itself in a situation with no good options but to either pay outrageous prices or curtail load (i.e. load shedding of interruptible load or if load shedding doesn't solve the problem, blackouts).
I have seen a lot of power systems and been involved up close and personal in a lot of electric power load forecasting, but I am constantly amazed that California doesn't have a better forecasting program. To be surprised on short notice with a 3,000 MW error is like needing to find 3 large nuclear power plants just sitting around doing nothing that you can call on for you forecasting errors. That is a heck of a lot of extra generating reserves that the system to carry just to compensate for bad forecasting.
Roughly five years ago (prior to the "Classical" crisis) the FERC did an extensive report on the failures of the mid-term forecasting system. Their primary concern was that the system operators were more or less completely inept in the day-ahead area, and that because of that we'd find ourselves getting whipsawed by the suppliers when we had to make mayday runs on the spot market.
The FERC was sufficiently irate that they levied something like a million bucks in fines. The state just blew the whole thing off, until the situation that FERC predicted happened. Whereupon, our Governor immediately went to work blaming it on someone else.