Regarding predictors, you first have to differentiate if you want to predict a diagnosis disregarding the imperative and preceding causality or if you want to make the assumption that causality must exist preceding the diagnostic event, and therefore, take into account an initial event. If you disregard causality, your predictor might be completely confounded by pre-CFS but post-causal factors. Thus, study designs would have to delimit their selection method to a certain assumed subgroup of the disease. Infection certainly seems to be the most adequate subgroup. Since initial EBV-associated mononucleosis happens during childhood for the most time, children are suited as participants.
In 2014, a study has analyzed exactly this question (
10.1080/21642850.2013.869176). I recommend reading it yourself, especially the discussion. Stress did not show to have predictive value at all. The study has limitations, though. The first limitation that they point out, though, I don't see critically. If they tracked stress during mononucleosis, you could only associate it with mononucleosis and not recognize it as a predictive factor for the reason mentioned above since EBV could also induce stress via its pathology. Since durational infection severity basically could predict CFS, this would then even contribute to the observational bias in their self-assessed questionary.
Typically for CFS, the sample size is small. The study only covered 6-months while CFS often is only diagnosed after at least 6 months. But the 6 month period seems rather arbitrary to me, just to improve specificity because the medical systems are very slow with co-diagnoses due to patient overload. And even if co-diagnoses are found, this is no exclusion of CFS, especially within the subsequent period of an EBV infection.
Their discussion also takes position on another study on psychological stress predicting CFS, which was conducted on a subgroup of children who were referred to psychologists. Clearly, this must be a confounding factor. How can you expect not to find psychological stress to predict CFS outcome when you only look at predisposed cases? It's just another study trying to push a narrative by design. You can associate fatigue with the psyche within any disease group if you look at predisposed cases only.
Overall, there is sufficient data that EBV and mononucleosis are associated with CFS. This study also shows that disease severity is most likely predicting post-infectious CFS. This makes infection a very likely causality of CFS and not the psyche.
Regarding the question of why certain cases are more severe than others, I believe that HLA genotypes will be determined in the UK DecodeME study to also show predictive value. I believe so because genotypes have been found for almost any other reactive autoimmunity, and infection susceptibility, especially for long-haulers and latent infections such as COVID-19 and antibiotics-resistant Lyme Borreliosis, which is HLA-DR4/DRB1-associated.
Some studies still might keep following the psyche narrative. Even if they find something, I'm convinced that it might only come down to long-term cortisol levels as the driving factor, which contributes to EBV and COVID-19 disease severity and duration. Yes, an association could be made between the psyche and CFS, but cortisol levels are mostly due to environmental factors such as lifestyle and workload and not due to any kind of trauma. Of course, diseases that are associated with cortisol levels, such as PTSD, will represent themselves with greater saturation in the overall population. But without an infection, they might very well never develop CFS.
This is just the nature of modern psychiatry. Everyone must have a psychological disorder and if they show none, a psychiatric disease must be involved. You probably know different anecdotes.