Atomicity means: the whole transaction settles or none of it does. There is no in-between state where you sent funds but did not receive them. In cross-chain, atomicity is expensive and rare. Most bridges are eventual: your source funds move first, destination funds follow.
Truly atomic: HTLC
Hashed Time-Locked Contracts. Both parties lock funds. Only reveal the secret to unlock both. If either party goes silent, funds return. This is how atomic swaps worked in the 2018-2020 experiments (bisq, Komodo). Slow, complex, mostly abandoned for scale.
Eventual: everything else
Canonical bridges, intent-based bridges, LayerZero-style messaging: user commits on source, some off-chain or on-chain process results in destination minting. There is a window where the user has sent source but not received destination. If the bridge fails, user has to file support tickets.
What can go wrong
- Refund needed. Some canonical bridges auto-refund after N days if destination cannot mint. Others don't.
- Stuck funds. Multichain 2023 froze $126M in transit when keys were compromised.
- Wrong destination. User typed wrong address on chain B. Bridge cannot recover.
What atomicity would fix
A truly atomic bridge would guarantee "you either get X on chain B or you keep your funds on chain A." No stuck funds, no manual refunds. Cost: complexity, latency (must wait for finality on both).
Where atomicity matters
- Very high value transfers where a failed bridge is unacceptable.
- Institutional flows where clear settlement is required.
Where it doesn't
Retail bridging $500 to a new chain. Eventual + good customer support is enough.
Koinlytics