This kinda belies the whole premise behind "smart" "contracts", doesn't it? I've always heard that the point of smart contracts is that we don't need the legal system to settle issues, because the code itself enforces the contract.

Apparently code can have bugs, and people can make mistakes? Whoops, wish we'd known those things before pouring millions into this stuff.

The intention of the parties was that the smart contract was enforced exactly as the smart contract was written: if that turns out to be unexpectedly disadvantageous for a party, then that's their problem.

I wonder if the argument will become what is the contact - essentially, is it specific transfers as implemented by the smart contract, or is it to be bound by the smart contract, even if these transfers weren't in the contemplation of the parties at formation?

This assumes a contract in addition to the smart contract. I am not sure in this case whether one could be said to exist, but certainly a truly decentralised system probably ought not to create such contracts.

The user (supposedly) read the (smart) contract and agreed to it as it is. So how is that resolved then?

Has a case like this ever been brought to court yet?

smart contracts are actually legally contracts (nor all that smart.)