well, forking and subsequent reorgs are local events, so we cannot be sure forkmonitor offers a statistical reliable view from global/overall perspective. But agreeing it does (I'm sure actual tuning cause few forks/reorgs because of the 10sec/10min ratio), my point is that long propagation times due to offline nodes, only sporadically peering with the BTC network, would increase the risk.
Disconnected nodes/client cannot be sure to share an updated status of the ledger, and if they are enough to have miners between them, well.. enjoy the mess   ...the matter is that BTC has a global consensus model, but the way it works depends on quite strictly bounded propagation times (which cannot be taken for granted with a significant number of offline nodes)