Before you can investigate the causes of an illness, you have to define it. Otherwise, you’re chasing an ever-shifting cloud of ambiguous symptoms, any of which could have different causes. The article opens with this admission, so I’m not stating anything new here.

The problem with “Long Covid” as it exists today is that there’s no such definition. Literally anyone who had Covid once and feels bad today (and quite a few people who never had a confirmed case at all) includes their set of symptoms in the communal diagnosis. Thus, if you dig into these studies, you always find that the syndrome is a wide-ranging and variable constellation of symptoms, making it impossible for a study to have any systematic legitimacy. Moreover, the results of any particular study are more strongly influenced by the inclusion criterion (if there even is one) than by any other factor.

It’s perfectly possible to evaluate treatments in this situation, and would be a better use of resources - pick symptoms, make an inclusion criteria, and run a randomized trial of existing drugs or therapies. But this is likely to fail, and it’s much, much easier to write papers with unprovable theories and retrospective analysis.