Most low-fire bodies contain talc. It is added for the express purpose of increasing thermal expansion. The natural quartz particles present do the same. These are good for glaze fit but bad for ware like this. There are also sudden volume changes associated with cristobalite, but it forms (from quartz) at stoneware temperatures so should not be a concern in terra cotta. You could fiddle with the clay recipe or change bodies, but better to change the firing schedule. While stoneware dunting happens between 950-1150F on the way down, this could be happening anywhere. A simple fix is to slow down the entire cooling cycle. Learn to program your kiln. Use a conservative cooling rate of about 200F/hr (even slower at 1150-950F). No electronic controller? Learn a switch-setting-schedule to approximate this down-ramp (buy a pyrometer if needed).
I document programs in my account at insight-live.com, then print them out and enter them into the controller. This controller can hold six, it calls them Users. The one I last edited is the one that runs when I press "Start". When I press the "Enter Program" button it asks which User: I key in "2" (for my cone 6 lab tests). It asks how many segments: I press Enter to accept the 3 (remember, I am editing the program). After that it asks questions about each step (rows 2, 3, 4): the Ramp "rA" (degrees F/hr), the Temperature to go to (°F) to and the Hold time in minutes (HLdx). In this program I am heating at 300F/hr to 240F and holding 60 minutes, then 400/hr to 2095 and holding zero minutes, then at 108/hr to 2195 and holding 10 minutes. The last step is to set a temperature where an alarm should start sounding (I set 9999 so it will never sound). When complete it reads "Idle". Then I press the "Start" button to begin. If I want to change it I press the "Stop" button. Those ten other buttons? Don't use them, automatic firing is not accurate. One more thing: If it is not responding to "Enter Program" press the Stop button first.
Here is an example of our lab firing schedule for cone 10 oxidation (which the cone-fire mode does not do correctly). To actually go to cone 10 we need to manually create a program that fires higher than the built in cone-fire one. Determining how high to go is a matter repeated firings verified using a self supporting cone (regular cones are not accurate). In our lab we keep notes in the schedule record in our account at insight-live.com. And we have a chart on the wall showing the latest temperature for each of the cones we fire to. What about cone 6? Controllers fire it to 2235, we put down a cone at 2200!
Dunting generally refers to firing cracks that occur in ceramic ware as it is cooled in the kiln. The reason is generally uneven cross section or too rapid cooling.
In ceramics, this refers to the sudden volume change in crystalline quartz particles experience as they pass up and down through 573C. Fired cracks are often related to this.
In ceramics, cristobalite is a form (polymorph) of silica. During firing quartz particles in porcelain can convert to cristobalite. This has implications on the thermal expansion of the fired matrix.