The Imperial College report has a lot of flaws. First, it runs the simulation largely using data from Italy, which is the worst-case scenario.
Second, the "18 months" is only one scenario they discuss.
Third, while counting for vaccines, it largely ignores any possibility of treatments arising (such as chloroquine, which is being used effectively in China and South Korea).
Fourth, it does not consider the "contact tracing" whereby, after ~3 weeks of social isolation, society starts realizing who is sick and they begin quarantining themselves, which is what we are seeing in South Korea.
Fifth, they do not consider how cases spread. I.E., most cases of identified COVID-19 are not air-spread, but by coming in contact with someone's germs. This happens frequently because of the long-time with asymptomatic illness. BUT, this means social isolation has a benefit that the Imperial College study does not address: those germs become contained.
Sixth, it barely accounts for a possibility where militaries are mobilized to build beds and other hospital infrastructure.
Seventh, these were the same guys that advocated for the herd-strategy until they realized the potential damage, then they switched their policy recommendation.
A link to a review that addresses the fourth and fifth points, specifically.
Listen, this will suck, it really will. Lots of people will die. But this Imperial College study is dangerous in that it ignores ergodicity in their simulation.