A possibly crazy EDS idea and perhaps a less crazy one...
I'm not sure how well the Thermo software fits EDS backgrounds, but on Bruker some materials don't result in great fits and therefore the background subtraction for net intensities can be off a bit. As an example, I have seen >0.5 wt% Al2O3 reported in a material with no Al because the background model was too low. Other than better EDS background fits, I wonder if MAN for EDS might work. To do this, one would need total counts within a ROI, not just the net counts. Yes, this appears crazy since EDS records the (sometimes cluttered) adjacent backgrounds, but as with the example above, there can be problems with these.
The perhaps more sane idea relates to the overhead of processing runs with lots of EDS data, an overhead that would be multiplied many times over if TDI using EDS intensities is implemented. Perhaps once they are first extracted, the EDS intensities could be cached in the MDB. Then, reprocessing would go much faster. Of course, there would have to be a way to repopulate the cache as needed, either on demand, or if one changes the EDS elements.