This analysis in Current Science was a nice read.
But it was confusing on one point. According to Table 1, only 12 repositories added at least one item per month for the one year from June 2015 to July 2016. But the same table says that 26 added more than 50 a month. This, I assume, means that 14 (26 minus 12) had at least one month in which they did not add a single item to the repository. That doesn't seem like such a bad thing. Regularity isn't such a strict moral virtue, is it?
Interestingly, if you exclude ICAR, the two biggest repositories by far are the Gokhale Institute of Politics and Economics, Pune and IIM, Ahmedabad. Then, comes IISc, North Eastern Hill University and IIT Bombay. A casual perusal through the Gokhale Institute's IR shows that they've uploaded digitized versions of old books like the Madras Records from Fort St George. I haven't explored the collection properly but there are surely very exciting opportunities for analysis once such records are digitized.