The question on everyone's mind is "when can we end the lock-downs" and "what happens if we unlock too early"?
This is a study by Lipsitch and Grad at the Harvard Center for Communicable Disease Dynamics.
They used epidemiological models to simulate various outcomes for varying lock-down periods, and simulated the "2nd wave" outcomes of various strategies. Their models take into account natural attentions of R0 during the summer months, varying amounts of attenuation from the various lock-downs, and other relevant factors. Simulations are provided for cases where interventions reduce R0 by 20%, by 40%, by 60%, and for the case where no interventions occur. They also produce a set of charts for "intermittent distancing" where there are multiple, short(ish) periods of social distancing.
I know everyone is going to ask me to summarize the results, but the very best way to understand the likely outcomes is to look at the curves in the paper. Page 9 contains charts that estimate prevalence and critical cases using the "intermittent social distancing periods" strategy. Graphs B and D on that page are (I think), the two best options. Graph B assumes the present level of critical care hospital beds in the US, and graph D assumes we manage to double the number of existing critical care beds.
If you're a visual learner, this is your paper :-)
This is a pre-print so the usually precautions apply. (i.e., not yet peer reviewed, may never get accepted for publication, etc, etc).
My personal bias is that papers that comes out of the HCCDD and the Harvard School of Public Health are usually of fairly high quality, but that is strictly a personal bias.
This may help give some insight to the consequences of removing the lock-downs at various times.