As someone who has worked in the Medical publishing world for 25+ years on some of the worlds leading medical journals you need to look at the status of the publication it's being published in, a study consisting of 84 subjects isn't anywhere near enough to be taken seriously results wise by and serious medical journal, it might be an interesting starting point that could start a discussion or kickstart a serious study, but nothing more than that.
Agreed. Those stats look far to good although that my be my unfamiliarity with the technique used, Cox’s proportional hazards regression analyses. Not an analysis I'm familiar with as I'm more familiar with scientific studies where you have before and after (plus control) actual measurements and you look for statistical differences (or similarities) between the groups. This is statistical modeling rather than statistical analysis of experimental findings I believe. My (admittedly limited) understanding of hazard regression analysis is that it requires that the hazards across the group must be the same except for the target factor (COVID) and constant over time. If for example those with COVID had trained far less on average than those who hadn't had it then that would be a violation of the first assumption and presumably, unless you had long COVID (I assume none did in the Belgian study as it wasn't mentioned) the risk isn't constant over time, as any muscle damage/effect would surely be more pronounced closer to the time of the infection than later?