Bitcoin, and probably any other cryptocurrency, uses flooding [1] to propagate transactions.

So it would be very difficult to prevent transactions from leaking out of or into the isolated zone:

A single peer in the whole country which is connected to the outside would be enough to leak all transactions.

[1] https://en.wikipedia.org/wiki/Flooding_(computer_networking)

How do you prevent double spend in the intervening period?

the risk is that a transaction you've accepted gets invalidated by a competing spend of its inputs in a chain coming out of a different "network zone"

so you wait for more block confirmations

If it is a valid threat model I don't know, but there are people streaming Bitcoin blocks over satellite for this very reason. So it is likely two islands of the network would be reconnected, provided that any one node is relaying blocks in this way.

I'd really like to see a serious analysis about the viability of a government, or an alliance of governments, to block / split the btc blockchain.

I find it hard to believe the US couldn't really do it if it wanted to, or even China (since most btc is mined there), but details like these make me wonder. Does anyone have any interesting material to read regarding this?

blocks can be broadcast over any communications medium. the shared state of the ledger would be smuggled in via any means necessary, from radio to messenger pigeon.

this remains true if retail payments migrate to second layer solutions. the important thing is sharing the same base layer anchor.

block confirmations would have to increase to mitigate the likely orphaning of chains coming out of different zones.

it's a good example of why achieving its civilisational infrastructure goals requires lightweight blocks.