~/notes / consensus and blocks
What Happens During a Chain Split
When part of the network refuses an upgrade, two chains exist and every holder owns coins on both. The mechanics are stranger than the summaries suggest.
A chain split happens when nodes disagree about the rules and neither side backs down. Both chains continue, sharing history up to the split and diverging afterwards.
Soft fork versus hard fork
A soft fork tightens the rules. Blocks valid under the new rules remain valid under the old ones. Nodes that do not upgrade continue to follow the chain, accepting blocks they would not themselves have produced. No split occurs if a majority of block producers adopt it.
A hard fork loosens or changes the rules. Blocks valid under the new rules are invalid under the old ones. Every node must upgrade or be left behind. If a meaningful minority refuses, two chains result.
Most planned upgrades are hard forks executed with near-universal agreement, and no split occurs because nobody continues on the old rules.
What holders own afterwards
Both. Balances existed on the shared history, so an address holds the same balance on each chain from the moment of the split.
Spending on one chain does not spend on the other. They are separate networks that happen to share a past.
Replay protection
The critical technical detail. If both chains accept transactions in the same format, a transaction broadcast on one can be replayed on the other, spending funds the owner intended to keep.
A competently executed split includes replay protection, typically a chain identifier that makes transactions valid on only one side. Splits without it have cost people money.
If a split is announced without replay protection, the correct behaviour is to move nothing until the situation is resolved.
What actually determines which chain survives
Not the merits of the argument.
Economic activity. Which chain exchanges list, which one merchants and applications use, which one holds liquidity.
Security. Hash power or stake securing each side. A chain with a small share of the original security is cheap to attack, and several minority chains have suffered successful double spends.
Infrastructure. Wallet support, node software maintenance, block explorers.
The name and ticker. Decided in practice by exchanges and data providers, and it matters more than it should.
The historical examples
The 2016 Ethereum split, over whether to reverse an exploit, produced two chains that both persist. The 2017 Bitcoin split, over block size, produced a minority chain that has since declined substantially.
In both cases the outcome was determined by where economic activity went, over months, not by the technical argument.
What to do as a holder
Before an announced split: move nothing during the window around it. Confirm your wallet supports both chains or at least the one you intend to follow.
If you self-custody: you control keys on both chains automatically. Claiming the minority asset requires wallet support and, if replay protection is weak, careful sequencing.
If you hold at a venue: the venue decides whether to credit the minority asset and when withdrawals reopen. Platforms publish their fork policies in advance. Venues serving retail customers, such as platforms that let you take custody immediately, typically state whether they will support both chains and what the suspension window will be, which is the only information a holder actually needs.
The broader point
Splits are the mechanism by which disagreements about rules get resolved in a system with no authority. They are messy, expensive, and the only available answer.
That they are rare is evidence the social layer usually works, not evidence that the protocol prevents them.
# Corrections and technical nitpicks are welcome. Send them over. They get published with the fix.