Keith got it right.
The update rate must be fast relative to the process. Sleepy Wombat's spread sheet and my web page used Ron's Beaufort's 'hot rod' gain and time constants. These were in minutes so updating every second was fast enough. Update every 6 seconds was a little coarse. Updating faster that one second was probably a waste of CPU. However, you should play with Sleepy Wombat's spreadsheet to prove this to your self.
Norm, unless there is some reason to keep some ideal fuel/air ratio constant, I don't think you need to ensure the flame goes completely off or on with SMC. I bet the flame stays always on unless the flame can consume all the gas each second or the valve is off for many seconds. The flame burning the gas will just be another time constant. In any case the on-off control is adding energy to the system in the form gas which the flame converts to heat with a little lag so controlling the rate of gas or time is what I think would be important.
I can add another time constant to my SMC control web sheet, but I have no idea what kind of time constant to use. Basically I would need to know how long after the valve goes of the flame goes off. If I consider this time to be 4 time constants then I think I can do a pretty good simulation with 3 time constants for heater, sensor and load.
OT
Norm, I know you are in the Chicago area and are intested in hydraulics. I will be at the DesignEngineeringShow on the 25th of Feb. I will probably be there for a couple days as we have a booth there. I will also be giving a presentation ( #332.2 ) on tuning and the steps required to do autotuning. It will cover how one finds the gains and time constants. A little knowledge of FFTs, poles, zeros, s domain and z domain would be good. I will be using data from our hydraulic system as an example. There will be a lot of hard engineering stuff with little fluff.