The unpredictability comes mostly from inability to measure/estimate the exertion across different activities. If you take just one exercise, walking for example, you can more reliably trigger, or avoid, PEM by walking certain distance/speed. That is the reason why patients employ fixed routine as a way to avoid PEM.
Adding to the anecdotes here: stomach issued had me make 5 trips to the bathroom (20m total) in about one hour. No additional pain or unexpected symptoms yet.
I think rolling PEM is an issue , particularly for new / mildly affected .
Some may. But most would agree that there is a strong relationship between exertion and crash most of the time. (There are triggers other than exertion that could trigger PEM seemingly out of the blue, but that does not mean exertion above certain level is a cause). The question here is whether the exertion threshold for the trigger exist. There is a way to test it if that is in doubt: remove other possible triggers best you can, let the patient rest for 3 days, and then have the patient walk certain distance at certain speed. If you can trigger PEM at or above certain distance/speed most of time, then we could conclude that exertional threshold exists. (Again, an observational study with actimeters could substitute). I think it is important that we establish that if you have any doubt about it. It's going to go a long way to accept/reject candidate hypotheses.
Actually, (total) exertion itself is rather easy to measure. You can define it as the rate of energy expenditure, or power/MET. It is the damage pertinent to PEM that is hard to measure. Smaller exertion of muscles that haven't been used for a while may trigger PEM while larger exertion of the walking muscles may not, for example. You need take into account the condition the patient is in at the time of exertion if you want to be precise. But establishing a general correlation between exertion and PEM probably is enough. That it is hard does not mean that it is irrelevant. though. That is why I'm working on it!
One thing I forgot to mention: patients in mild/recovered stage are less likely talk of crashes coming out of the blue. Other triggers and accumulation effect are minimal in that stage and therefore the effect of exertion is much more pronounced. This is why I keep saying that the research should consider the entire spectrum rather than focusing on severe/moderate side only.
This is true of my experience. When I was mild and still working it was very clear what my triggers were even with delayed responses.
Yes, some crashes come out of the blue, but that does not mean all crashes are totally unpredictable. Most of us know from experience that some specific exertions or stimuli, and the cumulative effect of combining some types and duration of activities, will always trigger a crash. That's the whole point of pacing - to reduce the frequency and severity of the predictable crashes when possible. We also hope it will reduce the likelihood of the unpredictable ones.
If you block the signals to/from the brain, how can you know that the brain isn’t keeping track in some way just because you don’t get the same results?
Or, the more parsimonious explanation would be that this activity has a pretty consistent impact on metabolic state, which is what the brain is actually sensing.
The brain is of course involved in received and integrating those signals, but the point @Snow Leopard is making is that any "accounting" the brain does is primarily a matter of sorting out sensory input. If it was solely a predictive model in the brain, you should be able to get accurate results regardless of whether or not the brain has access to that sensory input from metabolic-sensing afferents. Which is not what we see.
I must admit I probably use it a bit too often, but it's just so fun to say!
But accounting isn’t prediction. JE (to my understanding) is talking about keeping track of how much activity you’ve done. Where does the prediction come into play?
I think the idea is whether the brain is keeping track of how much activity you've done without any metabolic sensory input, which is what is meant by "predictive modeling" and "accounting." I.e. the idea that the brain accountant creates a "predictive model" of how tired you should feel based on how many steps it knows you walked (from the visual or proprioceptive systems, I suppose?)