On the other hand, most heat transfer relations are proportional to delta-T, so a T' = 1/T definition would make thing like q = h (T_2-T_1) become q = h( 1/T

On the other hand, most heat transfer relations are proportional to delta-T, so a T' = 1/T definition would make thing like q = h (T_2-T_1) become q = h( 1/T'_2 - 1/T'_1 ) Perhaps unlike the case for pi, I think there's a good argument here for keeping temperature as-is.