It isn't clear what the ETC does when normalizing the brightness of extended sources. Some experimenting makes me think it is probably setting the integrated flux of the extended source to match the requested normalization.
For the applications I'm considering, it would be much more convenient if the normalization could be specified in surface-brightness units. 1) Predicted or measured units for extended sources are naturally in surface brightness. 2) In the existing implementation (if my guess above is right), the effect of an extended source on count rates, SNR, saturation, ..., depends on the dimensions one assigns to that source.
This would be complicated to implement in the existing interface because the renormalization tab doesn't know what kind of source will be selected. That might be solved by moving the 'pt source' or 'extended source' button from the 'Shape' tab to the renormalization tab.