But I think, chances are, that new materials for permanent magnets will appear, that have somewhat better temperature coefficients. I think the research is going on, if not because of magnetic watch balances, the because of traveling wave tubes (radio frequency amplifiers).
Could that be that on the watch winder the main spring is constantly (almost) fully wound, therefore it has more power and the balance runs at full amplitude, i.e. the period is longer and the watch runs slower? It depends how the watch was regulated, if it was regulated for let's say 75% wound main
I think it is based on patent US7396154 by TH from 2007 (but the original application is even somewhat older) . You can find it on freepatentsonline for example. There are interesting diagrams and more explanation. It also refers to some other simmilar concepts. Also temperature compensation is disc
COSC criteria is 0,6 sec/day/ 0 C, which is 0.6/24/3600=~7E-6, i.e. 0.0007% change of period due to change in temperature of 1 0 C. The best current sammarium cobalt permanent magnets have RTC 0.002%, i.e. a change in magnetic flux Br per degree Centigrade is 0.002%. on the range -50 +150 0 C, which