Take this paper with a huge grain of salt. They went on a fishing expedition, then pointed an ML model at the data and reported some clusters. This is...questionable methodology. ML models will find a way to draw a discriminatory boundary for almost any classification problem.
On the other hand, there was this paper [1][2] from the NIH, which was a well-controlled, pre-registered cohort study, did pretty much every clinical test imaginable (about 150 different measurements per person), and found no measurable differences between "long covid" people and controls. This is a much better study, and rebuts the claim.
Let's be 100% clear: this is all speculation. I take no side on any of it, except to say that the evidence is bad, and what little good evidence there is doesn't support the hyperbolic claims of prevalence, nor does it support the claims of any particular mechanism or theory of what "long covid" is (if anything at all).
On the other hand, there was this paper [1][2] from the NIH, which was a well-controlled, pre-registered cohort study, did pretty much every clinical test imaginable (about 150 different measurements per person), and found no measurable differences between "long covid" people and controls. This is a much better study, and rebuts the claim.
[1] https://www.medpagetoday.com/infectiousdisease/covid19/98885...
[2] https://www.acpjournals.org/doi/10.7326/M21-4905