Cross-chain messages, and why they confuse people
A bridge transaction is not one transaction. It is a message sent on one chain and consumed on another, and the two halves have different hashes, different timestamps and different explorers.
This is the root of nearly every bridging panic. You deposit on Ethereum, see a confirmed transaction, and then find nothing on Arbitrum because you are searching for the L1 hash on the L2 chain, where it does not exist. The funds are fine. The identifier is wrong.
Arbiscan's message tracker resolves this. Give it either side and it shows the pair — the L1 transaction that initiated the deposit and the L2 transaction that credited you, or the L2 withdrawal and its eventual L1 claim. It also shows the state in between, which is where a message can legitimately sit for some time.
Deposits are quick, typically around ten minutes, because crediting on L2 does not require any dispute period. Withdrawals are the slow direction, for the same structural reason as Base: an optimistic rollup must allow time for fraud proofs before releasing funds on Ethereum.
The seven-day wait
Canonical withdrawals from Arbitrum to Ethereum take roughly seven days. This is not a queue, a backlog or a performance problem. It is the challenge window, and it is the security model working as designed.
The sequence: you submit a withdrawal on Arbitrum, the state root containing it is published to Ethereum, the challenge period runs, and then you submit a second transaction on Ethereum to claim. That final step is yours to make. A surprising number of people initiate a withdrawal, wait a week, and never claim — the funds sit unclaimed indefinitely, recoverable, but only once someone realises.
Arbiscan shows which stage a withdrawal is at, which is the only reliable way to tell "waiting normally" from "waiting for you to do something".
Worth knowing
Fast bridges avoid the wait by giving you funds on Ethereum immediately from their own liquidity and absorbing the seven days themselves. That is a real service and they charge for it. What you are buying is time, and what you are accepting is a counterparty — which the canonical bridge does not have. Neither choice is wrong; they are different risks.
Arbitrum gas is two costs pretending to be one
An Arbitrum transaction fee has two components. There is the L2 execution cost, which is tiny, and there is the L1 data cost — your share of publishing the transaction data to Ethereum — which dominates.
This produces behaviour that looks wrong if you expect Ethereum's model. Fees on Arbitrum move with Ethereum congestion rather than with Arbitrum activity. A quiet period on Arbitrum during a busy period on Ethereum means higher fees, which makes no sense until you know where the cost comes from.
Since EIP-4844 introduced blob transactions the L1 data component became substantially cheaper, which is why rollup fees fell sharply across the board. Arbiscan shows the total; it does not break out the split, which is a genuine gap when you are trying to understand why a transaction cost what it did.
Arbitrum Nova, briefly
Nova is a separate chain using AnyTrust rather than a full rollup, which means transaction data is held by a data availability committee instead of being posted to Ethereum in full. That makes it cheaper and moves the trust assumption from "Ethereum has the data" to "the committee is honest".
It has found a niche in gaming and social applications where fees matter more than maximal security. Arbiscan covers it, though visibly as a secondary concern — less tooling, fewer verified contracts, thinner coverage generally.
If you are using Nova, know which chain you are on. The two have separate explorers, separate bridges and separate contract addresses, and a contract deployed on One does not exist on Nova.
Verdict
If you bridge to or from Arbitrum, use Arbiscan. The message tracker is the feature and nothing else replicates it.
For ordinary contract work it is Etherscan with different data, which is fine and entirely predictable. Blockscout has Arbitrum coverage if you want open source, and 3xpl is faster for a quick lookup.
The one thing to take away is the same one as every rollup: confirmed on L2 and settled on Ethereum are different events with different guarantees, and the interface uses one word for both.
What our reference tests showed
Our Arbitrum testing concentrated on the bridge, since that is what the message tracker exists for.
We deposited from Ethereum and then searched for the L1 transaction hash on Arbiscan, reproducing the mistake people actually make. It returned nothing, as it must — the hash does not exist on L2. The message tracker resolved it immediately, showing the L1 transaction and the L2 transaction that credited the funds about ten minutes later, linked as one movement.
We then initiated a withdrawal and followed it. Arbiscan showed it move from submitted, to included in a state root published on Ethereum, to claimable. The final claim required a separate Ethereum transaction that does not happen automatically, and the explorer made clear that it was waiting on us rather than on the protocol. That distinction is the one people most often miss.
The gas breakdown was the visible gap. Fees on Arbitrum are dominated by the L1 data cost, and the explorer shows a total without separating the two components — so a fee that rose because Ethereum was busy looks unexplained.