[ghthor]

One idea I have is to execute health checks to the network from different points in ipfs node graph. For example you could add your file to your localnode(connected) then spawn some geographically distributed ephemeral modes to request the file. Once all the your emphermal nodes have recieved they wipe and you take your localnode(offline). Now the ephemeral nodes heartbeat, at some interval t, to request the file with the original source offline. If some N of them don't receive the file in some timeout, reconnect your localnode. In theory this would keep your file alive within the network for some x*t timespan, the question is what is that timespan and what's the cost of running the heartbeats.... I'm actually going to run this experiment and I'll reply here with my results.