I can't comment on T23- but there are to be 2 conferences on this subject this year ( April 09 near Zurich, and June 09 in Louisville)
regarding P91 cooling rates, it is suggested one goes back in history and re-reads the final report by ORNL ( V Sikkha) on P91 optimization. There, one finds several TTT curves for the purpose of optimizing normalization times and temperatures and cooling rates. Some highlights include:
a) normalization time and temp affect crystal size, which in turn affects creep strength and also the propensity to form ferrite upon cooling; a high normalization temp yields large crystals, low creep strength, and lower propensity to form ferrite upon cooling. I am not sure if modern solid state thermodynamics programs reflect this trend. But I have seen foundries use a relatively high normalization temp ( higher than ORNL recommends) plus slow cooling rates- I would expect lower than optimum creep strength with that combo.
b) a slow cooling rate can yield high ferrite content if the cooling rate is constrained in the temp range 1800-1400 F- as seen in the Mannesman valouurec P91 booklet for a specific assumed normalization time and temp batch
b)It is known ( from similar measurenment error associated with boiler fluegas measurements) that an error on the order of 100F can be expected at the monitoring thermocouple attached to the workpiece being normalized at 1900 F, due to radiation of heat to the colder enclosure- so the means of attaching and shielding the T/C to the workpiece will affect the accuracy of the monitoried and controlled normalizing temp, and the resulting crystal size and metallurgy.