Getting a bit OT for this thread, so I'll try to be brief: Iceland has a company named deCODE Genetics, which was designed around doing whole-population genetic studies for disease modeling, so the company has been repurposed into doing randomized sampling of the population to look for COVID-19 in people who are not suspect cases - just the general public. They found the disease in around 1% of the general population over a period of a couple weeks (reported 1 week ago) - an implied per-capita rate twentyfold that of the country's official diagnoses at the time of reporting (which in turn has involved one of the most aggressive testing regimes in the world). Over half had no symptoms whatsoever, and the rest, generally just a minor cold.
In an experiment which started several weeks ago, Italy tested the entire town of Vò in a series of rounds and successfully eradicated the disease from the town. Again, this is one of the few cases of experiments of testing of the general population rather than suspect cases. 3% of the population was determined as infected during the first round. 50-75% of them were entirely asymptomatic. Most of the rest had only minor cold symptoms and did not suspect that they had the disease. During the study, asymptomatic cases usually remained asymptomatic, and most minor cases remained minor.
The net picture is that the disease was already widespread even several weeks ago; that people with no symptoms whatsoever are the majority; and of those with symptoms, symptoms so minor that people don't get tested are by far the most common. To match with the death rate (as severe cases almost always get tested), a widespread disease in the population implies a low IFR (infection fatality rate), and implies that places with overloaded medical systems are overloaded not because the disease sends a large percent of its victims to the hospital, but because a large percent of the population has already gotten infected. This implies that the disease will tend to play itself out sooner.
As mentioned, the other study I mentioned (still preliminary) estimated that Wuhan hit a 19% infection rate at its peak, and that the disease had a non-time-delay IFR of 0,04% and a time-delay IFR of 0,12%. They also found (as others have) that the disease is highly prone to nosocomical transmission (spread inside hospitals and clinics, generally by staff) - that is to say, prone to spread to the most vulnerable of patients. I.e. that stamping out nosocomical transmission is one of the most important steps that can be done. Both Wuhan and Lombardy had extensive problems with nosocomical transmission before the severity of the situation was realized.
ED: Well, so much for trying to be brief...