When try, try, try again leads to out-of-order execution bugs
Summary: Trail of Bits researchers reviewing Arbitrum Nitro and contracts integrating with it found that retryable tickets, which carry calls from the Ethereum base chain to the rollup, are expected to execute in submission order but are not guaranteed to. Because failed tickets persist in memory for up to one week and can be redeemed manually by any user, transactions created together can execute out of order or in partial subsets, and the firm released a Slither detector named out-of-order-retryable to flag functions that create multiple such tickets.
Why it matters: A protocol that assumes its second queued transaction runs after the first can be broken or hacked when an attacker redeems them in reverse order, as in the post's example where a user's unstake executes before claim_rewards and their rewards are lost.
Read the original at Trail of Bits Blog ↗