Yes, of course there will be unobserved and uncontrollable factors in these types of observational studies but that does not necessarily invalidate the findings. It just speaks to the level of confidence. That’s why researchers should strive to control for as many observables as possible.
For example, one can compare two samples of men aged 65+ with diabetes and a bmi over 30. There may still be some unobserved heterogeneity but the more subgroups that can be isolated showing a similar pattern between control and test groups the higher the level of confidence that the overall finding is not just a statistical anomaly.
But ultimately, this is just one study and others will need to be conducted with different populations and time frames to determine the long-term effects of COVID.