I think that is a useful point but bone marrow output of myeloid cells can respond pretty rapidly. I don't know details but after haemorrhage platelet numbers can go up within a few hours and I presume that is from marrow release. Similarly, with infection the neutrophil count can rise fivefold in a few hours.
I think the sort of mechanism proposed might be subject to overriding with acute signals much in the way that people talk of adrenaline overriding PEM in the very short term. Some also talk of getting better ME/CFS wise during infections. In RA we think there are underlying loops in immune memory but joint inflammation can appear and disappear over hours too.
I also think it is likely that any model like this might be one module of the process with neurological loops being other modules with some people's disease being almost entirely driven by immune loops and others' by brain loops but with both being involved to some degree in all.
If the immune loop is anything like the ones in RA and psoriasis (a B cell loop and a T cell loop) it seems they are consistent with just about any long term dynamic pattern in different cases. I think that may be an argument for invoking an adaptive clonal element but I am not sure. Just basing things on the rules of myeloid cells I would expect monozygotic twin concordance to be almost complete because they would meet relevant infections at some point and the pathways and regulatory steps engaged ought to be much the same.