It's not, though. Someone suffering from severe symptoms that aren't COVID-19 may have other treatment options available, and may be placed in hospitals with less strict isolation requirements. In a perfect world we would have isolation beds and supplies for everybody, but that still doesn't get around the fact that treatment for COVID-19 and bacterial pneumonia are two different beasts, and both lead to similar end results.
And someone suffering from severe symptoms has other diagnostic options: a chest X-ray or CT.
If your goal is to find people with coronavirus (and this should be your goal if you want to slow its spread) you'll almost always find more of them by testing people without symptoms.
Together, those two things tell me we should have such a large testing capacity that testing symptomatic people isn't a stretch. (Of course, we might need more reliable tests if their medical situation is shaky, which is the different category of testing capacity.)
We should, but we don't. Group testing dramatically expands the number of people we can test. However group testing collapses at high prevalence. Since we expect people with fever and cough to be more likely COVID+ than the general population prevalence when we can only run a limited number of cases we'll inevitably miss more positives by testing people with symptoms.
Testing people with symptoms is a waste of tests. When tests are in short supply assume everyone with symptoms is positive, isolate them and test their contacts who don't have symptoms. As of last week, Boston Medical Center had a 44% positive rate on the tests they were running on symptomatic people. They could find more positive cases per test by testing people without symptoms.
https://www.medrxiv.org/content/10.1101/2020.03.27.20043968v...