for me, it not so unthinkable that you could have intermittent access to the internet (out days at a time), or that some parts of the internet would not be accessible for extended periods of time
for you it seems that must equate the end of the world where it doesn't matter anyway.
A "rejoin" isn't really possible. People who spent money on the "wrong" chain will simply have to live with the fact that their money is lost on the "right" chain, although it will probably not be an overnight thing. The two forks could exist for a long time after the split simultaneously, and many merchants may accept either form of payment. There will even be arbitrage between the two chains as profit can be made.
Ultimately, one chain will come to dominate the other, and the ones who fight hardest for it to be their chain will probably find they're on the wrong one. That's because people on the right chain will not need to fight about it. They'll just be running their software as they've been doing till now, and it will pick the correct chain for them. The ones screaming "you need to download a new client" will proceed in destroying their own reputations.
The devolving of the internet to national intranets is likely, but I don't think it can be done to a big enough extent that it stops Bitcoin. It only takes information to leak through the firewall at one point of entry to make the firewall completely useless. Bitcoin transactions and blocks can be transmitted globally via sat, long and short distance radio, telephone, or the internet, bypassing the firewall through steganography and such. It seems like a futile task for states to attempt to filter it. The Bitcoin protocol will likely be available over encrypted and authenticated transport channels too, meaning that the nation state isn't going to be able to stop it without completely banning encrypted communication for which it has no back door.
That may be attempted, but then no technology company is going to start a business in the locations where this happens. The nations which attempt this will probably do more damage to their economies by enacting it than any damage that Bitcoin could possibly do. It will only take a couple of nations to declare themselves open for bitcoin business, to have economic success, for the rest of the world to realize that fighting against Bitcoin is a losing game.
Debasement of the USD will probably happen, but who knows when. My guess is it that it won't be a sudden event but will happen gradually over a few decades. You might measure USD to lose half of its value over a decade, but if you attempt to do the same for bitcoin, you might find that it has barely changed because "inflation" in bitcoin is negligible - ~7/8 of the total bitcoin supply has already been released, and the amount of "inflation" is decreasing over time.
> have to live with the fact that their money is lost on the "right" chain
It's not that simple. Transactions can, and will be, replayed on both chains. There would however be many opportunities for attacks such as double spends for a short window of time.
> The two forks could exist for a long time after the split simultaneously,
That's not realistic. As soon as there is some communications between the chains, the system will reconcile. It's sort of the whole reason why this proof-of-work thing exists. Without the need to reconcile chain splits it would be sufficient with some sort of voting of proof-of-stake scheme.
> It's not that simple. Transactions can, and will be, replayed on both chains. There would however be many opportunities for attacks such as double spends for a short window of time.
My point was that a merchant who accepts a payment for some good or service, and later finds out that they were partitioned from the network after they have already released the goods, may find that they don't have any money on the other (main) chain because it was already spent. The merchant has no recourse to get back their money.
> That's not realistic. As soon as there is some communications between the chains, the system will reconcile. It's sort of the whole reason why this proof-of-work thing exists. Without the need to reconcile chain splits it would be sufficient with some sort of voting of proof-of-stake scheme.
This is what I was alluding to in the last paragraph. The majority of participants who do nothing will simply continue to operate on the main chain, but those who were partitioned (which may include miners) may resist operating on the main chain by forking their software. If enough people are affected by a partition, they would have the incentive to follow the partitioned chain and promote the forked software because their money doesn't exist on the main chain. They would probably operate both chains in attempt to spend the money they received whilst the partition occurred, but eventually the dominant main chain would account for most of their economic activity and the partitioned chain will become obscure, like Bitcoin Cash.
> they don't have any money on the other (main) chain because it was already spent
Right, the double spend attack. It's feasible under certain circumstances, such as a chain split. That doesn't mean the system breaks down completely, but that it has to be mitigated. In the event of a multi hour long network partition, some participants are likely to take action.
> they would have the incentive to follow the partitioned chain and promote the forked software because their money doesn't exist on the main chain
My point was that their money do exist on both chains as long as transactions are replayed. That may be more or less hard depending on the nature of the split (say, a whole country falls on the Internet completely). Economic participants do have an incentive to replay transactions (for example by following the satellite feed, or a number of other ways).
It is enough that one participant does this to at least give everyone else the possibility to mitigate themselves. Not a good situation to be in, of course, but still. It doesn't require nodes to follow both chains, and certainly not promote minority chains. There is simply no economic incentive to do that should transactions be guaranteed on both chains.
>That's not realistic. As soon as there is some communications between the chains, the system will reconcile.
If you think that is possible then you must not have paid attention to how blockchains work. By definition there is only one chain and it is always the longest chain in the network. The shorter chain will always be discarded and all the data within it is lost. Now we have a problem. Can't everyone just create their own chain with garbage data and then win by being the biggest spammer? (also known as 51% attack). This is why proof of work is necessary. It is basically a cryptographic lottery that ensures that there will only be one winner every 10 minutes. Since there will never be more than 1 winner there will also never be more than one longest chain. If you decide to disconnect from the network and solve the POW offline your chain will always be much shorter than the main chain. Only if you possess 51% of the miners can you do this offline attack and control which chain becomes the new one. It is possible to rewrite past transactions as well. Just start mining 3 blocks ago and remove transactions or add new ones. However, this becomes harder and harder the older the transaction is. Finding e.g. 10 blocks takes such a huge amount of mining resources that it is infeasible if don't control at least a super majority.