How well will Massive MIMO work at longer ranges (beyond 1000 metres)?

4G Massive MIMO gives a major boost to spectral efficiency in the 2.5 and 3.5GHz bands in Asia (China, Japan, S Korea) when used over ranges up to a few hundred metres.  European operators are seeing the same benefits in similar situations (at a relatively small number of sites).

For 5G, there's a widespread hypothesis that 3.5GHz M-MIMO antennas can be deployed at existing macrocell sites, and will work well out to a radius of one or two kilometres - with M-MIMO boosting the propagation in the downlink and the uplink, as well as giving higher spectral efficiency.  

Whilst this seems to make sense, I've not found any examples of M-MIMO delivering these benefits in practice in real networks.  In addition, a couple of vendor contacts have suggested that M-MIMO may not be very effective at longer ranges without some re-engineering.   

I'd love to hear any views on this topic.  Clearly if 3.5GHz 5G can be deployed at existing cell sites with good coverage areas, that makes a huge difference to the cost of deployment, and the overall 5G roadmap - at least in Europe.