Sectioning someone because they might be a danger to their children is pretty extreme. It might work as a temporary crisis measure, but it would more usually be dealt with by the family courts giving residence to the other parent (or put the child in care if there is no other suitable relative) and giving the parent with the personality disorder supervised access only. But it would have to be really bad to get to that stage and very few narcissists ever even get a diagnosis - they are able to give a good impression to doctors etc.
However as children get older the courts give them more say in their living arrangements and kids figure this stuff out in the end.
My mother was a narcissist. My father was completely under her thumb. I emerged from childhood with a completely warped view of the world and it has taken a lot of time, money and suffering to deal with it. I only ever had a few days alone with my father without her imposed itinerary, but I cherish those memories, and if they had split I hope I would have had more time like that in the normal world.. But it wasn't going to happen. I went around with a lot of guilt because I believed that my mother made all kinds of unbelievable sacrifices just for me, whereas actually everything was arranged to be about her in the end and I was being manipulated.
Good luck, and be aware that if there is any manipulation going on, the kids will benefit immeasurably from even a short break from it.
Could the Harvard guy be Prof Lieberman? His research page is here http://www.barefootrunning.fas.harvard.edu/
Personally I would be a bit concerned about running in a big chunky trainer in which the padding has lost its spring, because such shoes hold your foot in a heel striking position, and heel striking with inadequate padding will expose your foot to very high impact forces. I doubt that would be comfortable or healthy.
if you have a read of some of the entries in http://www.parkrun.com/crispy/graphicoftheweek
Someone has created an equivalence chart for parkruns based on times of runners who run two different parkrun courses on consecutive weeks (so on similar form). As parkruns run every week, if you have enough data points the effect of weather will cancel out. This metric will not disregard the effect of greater competition - but that is what you want if you want a race time prediction at a busier or quieter course. However it will assume that the competitive effect will be the same for all speeds of runners at a pair of courses.