Before the first vaccination was invented, for smallpox, there already was a technique called inoculation or https://en.wikipedia.org/wiki/Variolation. Unlike vaccination where dead or modified viruses are introduced to the immune system without causing an illness, in those old-style innoculations, people were infected with real live smallpox viruses, but not through the normal infection mechanism (the mucus membranes) but through scratches in the skin. Apparently when the virus arrives in the "wrong" organ, it can't cause as much damage quickly, but still the immune system gets a chance to prepare antibodies for it - so at the end the individual does get sick, but it's a much milder version and not as deadly.
Could we do something similar for COVID-19, i.e., deliberately infect at-risk individuals in locations far away from the pulmonary system (e.g., scratches in the skin) hoping such innoculation will indeed cause disease but a milder and less deadly version?
Is there any reason, besides the ethical questions of how to test this approach, why such an approach might not work? (again, "work" here means reducing the morbidity rate. It will not reduce the number of people sick, unlike a proper vaccine - it should just reduce the number of critically ill and most importantly, the death rate).