This is an interesting idea. One downside is that baselines from the Terahertz Array to the main ALMA array won't be so useful, and that if the atmosphere is very different cross calibration may not be quite so successful. To some extent, these depend on exact siting, so the degree of problem might be ameliorated. Al