[jacquesm]

Shutting down a nuclear reactor can be done very fast (emergency shutdown, underload shutdown) but increasing the load takes much longer.

This is one of the reasons that the blackout of 2003 lasted as long as it did, due the rapid decrease in load a couple of nuclear plants did a rapid shut-down.

Most base-load generating plants are difficult to adjust to a rapidly varying load.

[Retric]

You could set a bypass around the generators to the cooling tower to reduce energy production without reducing the rate of nuclear reactions. They don't do this for a number of reasons which basically boil down to nuclear fuel being cheap so they would much rather shut down just about everything else.

This does create a minor issue if the grid fails, but that’s a rare event and doing an actual shutdown is considered safer.

[jacquesm]

Ah, so that's the reason. I actually wondered about this after the 2003 event, I was caught right in the middle of it and had plenty of reason for reflection on the theme of power generation (that plus a lifelong interest in renewables also caused me to look into the various alternatives).

[logfromblammo]

"Neutron poisons" that accumulate in fuel rods during normal operation can also interfere with a re-start. Sometimes a shutdown core has to sit for a few weeks to let those reaction poisons decay before it even becomes possible to sustain a controlled chain reaction in that core again.