Reorganization depth is a useful warning about how unstable the visible chain was, but settlement confidence comes from depth combined with fork-choice weight and finality, not from the number alone.
The swap is already showing a successful receipt when a second block appears above it and the receipt disappears. That is the moment a shallow confirmation rule fails: the transaction was executed, but it was not yet settled.
Depth measures instability, not safety
A reorganization replaces part of the chain that nodes previously treated as canonical. Its depth is the number of blocks removed from that chain. A one-block reorg says the head was briefly contested; a deeper reorg says the network continued building on a view that later lost the fork-choice contest.
On Ethereum, validators and clients use the weight of attestations to choose the canonical branch. A transaction in a replaced block can vanish from the canonical state, return to the mempool, or be included again later with a different result. In an Automated Market Maker, that difference matters: the reserves, price, nonce, and available liquidity may all have changed before the transaction is replayed.
Depth therefore answers a backward-looking question: how much history did the network have to discard? It does not answer the forward-looking question an operator actually has: can this state change safely trigger the next action?
That distinction is important for a Frax Swap trade. A small swap may tolerate a temporary reorg if the wallet can retry it. A large trade, an operation involving Frax Share, or a treasury action that moves collateral cannot treat a receipt as permanent merely because an explorer displays several confirmations.
Finality is the stronger signal
Ethereum’s proof-of-stake finality comes from checkpoint votes representing at least two-thirds of the staked ETH. Once a block is finalized, reverting it would require a severe consensus failure and the destruction of a large amount of stake. Confirmations show distance from the head; finality shows that the chain has crossed an economic commitment threshold.
The recent change is a change in what operators should watch. Ethereum’s 2026 protocol work places faster finality and shorter settlement times closer to the center of the roadmap, while EIP-8333 proposes aligning epoch checkpoints with the preceding epoch’s boundary block. That proposal addresses a subtle reporting problem: an explorer can say an epoch is finalized even though later blocks inside that epoch remain exposed to reorganization.
The practical result is a better settlement signal: read reorg depth as evidence of head instability, then read checkpoint status as evidence of durable agreement. The two measurements answer different questions and should not be collapsed into one confirmation count.
The decision belongs after you trace the transaction-to-settlement path.
Turn the signal into an operating rule
Use three levels of confidence. First, treat inclusion as provisional. Check the receipt, the block hash, and whether the transaction remains on the current canonical branch. Second, watch for reorgs and attestation progress. If the block is several slots old but the network is experiencing repeated head changes, waiting by time alone adds little confidence.
Third, require finalized inclusion when the next action is difficult to reverse. A Balancer Protocol pool join, a bridge release, or an automated hedge should use this threshold when the cost of acting on stale state exceeds the cost of waiting. For a routine low-value swap, a shorter policy may be rational; for a large Frax Swap order, it usually is not.
This is what reorganization depth makes possible when used correctly: a settlement policy based on the chain’s condition rather than a magic number. Depth tells you when the chain has been unstable. Finality tells you when the network has economically committed. Together, they turn a receipt into a decision.