It has been previously observed that the SARS-CoV-2 virus was relatively stable at 1 mutation a month. However, recent studies, and the one you refer to suggest that this is incorrect.
Based on current data, it seems as though SARS-CoV-2 mutates much more slowly than the seasonal flu. Specifically, SARS-CoV-2 seems to have a mutation rate of less than 25 mutations per year, whereas the seasonal flu has a mutation rate of almost 50 mutations per year.
based on data stored at Gisaid
However, a paper just released on the 19th April 2020 lead by Li Lanjuan (it was on her advice that the city of Wuhan was quarantined ) based on 11 early samples taken from patients in Zhejiang at the beginning of the pandemic show otherwise. These patients had close travel links to Wuhan
Professor Li Lanjuan and her colleagues from Zhejiang University found within a small pool of patients many mutations not previously reported. These mutations included changes so rare that scientists had never considered they might occur.
Li’s team detected more than 30 mutations. Among them 19 mutations – or about 60 per cent – were new.
They found some of these mutations could lead to functional changes in the virus’ spike protein, a unique structure over the viral envelope enabling the coronavirus to bind with human cells. Computer simulation predicted that these mutations would increase its infectivity.
To verify the theory, Li and colleagues infected cells with strains carrying different mutations. The most aggressive strains could generate 270 times as much viral load as the weakest type. These strains also killed the cells the fastest.
It was an unexpected result from fewer than a dozen patients, “indicating that the true diversity of the viral strains is still largely underappreciated,” Li wrote in the paper.
So, different mutations were found in the same patients, and different degrees of lethality when tested against cell cultures.
They think this may explain why differing populations have been experiencing different degrees of disease severity across the world. These mutations had not been noticed before because the genomes published on worldwide databases were established using standard techniques
Most of these samples, though, were sequenced by a standard approach that could generate a result quickly. The genes were read just once, for instance, and there was room for mistakes.
But Li's team did deep sequencing to pick up these mutations
Li’s team used a more sophisticated method known as ultra-deep sequencing. Each building block of the virus genome was read more than 100 times, allowing the researchers to see changes that could have been overlooked by the conventional approach.
Patient-derived mutations impact pathogenicity of SARS-CoV-2