I have a similar situation although the size of the Moodle site in the cloud is much smaller. The cloud site is a MoodleCloud site. We have three schools in the Northern Region of Ghana running MoodleBox; one of them is just beginning so they have nothing to report, yet. The other two have only one Moodle course for 25 students each. Periodically, once every two months, so far, the teachers back of their courses and send me the course file which I then upload to the 'main' Moodle site on the MoodleCloud site. So far, the uploads are individual courses.
Our current problem is determining what reports to run from each course backup so that we can compare utilization from period to period. We are primarily interested in teacher activities on the course and student activities on the course.
This is not an automated system, but I'm fine with that. After we determine which data from the courses is most useful, we can focus on making the process of collecting the data and merging it with a larger database of teacher and student activity.