According to this article from Reuters, it takes up to eight hours for a coal plant to reach maximum generation. This means that four up to eight hours after it starts up, the plant is burning coal at a lower efficiency (Btu/calorie per kWh) than what it's capable of during longer-term operation.
In the U.S. cycling of coal plants is something that happens on a daily basis in some areas. The chart below (from the Energy Information Agency) shows power sources over an average day in the western U.S.:
For a huge area of the U.S., coal plants are reducing output for several hours each day. Presumably this involves turning off boilers or even whole plants.
This seems similar to the question on shutting off a car's engine at a stoplight. Could keeping a coal plant running, in some circumstances, result in less CO2 emissions than stopping and restarting it?
To figure that out, I'd need some data on how much coal a plant burns during shutdown and start up, and how much power is produced during this period, compared to how much coal is used to produce the same amount of power after it's been online for some time.
This question was prompted by the recent news that Britain went a week without burning coal. There's also a good discussion over on skeptics.se about a similar occurrence a few years ago: During Great Britain's coal-free day, were coal plants shut off, or active and generating power GB didn't use?