From reading the article it seems all that is needed is to have a more diverse and resilient set of bootstrap nodes.
Maybe ask trusted members of the community to run bootstrap nodes in different locations and on different cloud providers or even physical machines, then curate them into the official bootstrap list?
Also, persist peers over daemon restarts to have some peers in case all official bootstrap peers are unreachable? Can’t see anything wrong with that, and should be pretty easy to implement.
There’s an issue on the github repo for exactly this.
It could help to move the default bootstrapping list out of the configuration file too. That way, the list could be updated in future versions as better and more diverse infrastructure becomes available.
You really don’t need to use the bootstrap peers more than the first time you use IPFS. They’re only supposed to get you into the network after all. The network may temporarily split into separate networks, but this isn’t a real problem as it only takes one node that knows any peers between the different networks to rejoin them. You could mitigate this by having clients connect to one randomly selected default bootstrap peer every ten startup or something.
Anyhow, the important part is to get session persistence for DHT nodes implemented and get the ball rolling.