Hello Steven,
regarding the memory consumption during the alignment, I can see from your first screenshot that you used very high values for the detected features which consumes too much RAM. In this case please try to lower the number of detected features and try again.
Regarding the memory consumption during the reconstruction, it looks like bug on our side, we will take a look at it. In the meantime can you please try the workaround proposed in the article you have mentioned in your original post https://support.capturingreality.com/hc/en-us/articles/115001483771-Huge-scene-reconstruction ?
If you abort the reconstruction, then only depth-maps are reused but the model parts are calculated again.