Computer Occupations, All Other Interview Questions & Answers

Example answer

“I found a reentrancy path in a staking contract during a Foundry invariant campaign two days before the planned mainnet deployment. The withdrawal function transferred an ERC-20 token before decrementing the user's stake, so a malicious token callback could have withdrawn against the same balance repeatedly. I paused the release, reproduced the exploit on an Anvil fork, and estimated that the full launch liquidity pool of $1.8 million would have been exposed. I changed the flow to checks-effects-interactions, added ReentrancyGuard, and wrote five invariants covering total staked balance, reward debt, and withdrawal limits. The follow-up audit reported no critical or high-severity findings, and those invariants became required CI checks for every contract change.”

Example answer

“Our product team wanted every document revision stored directly on Ethereum because they equated on-chain data with trust. I mapped the proposal against cost, privacy, retrieval speed, and auditability, then showed that storing 10,000 monthly revisions on-chain would cost roughly $18,000 per month at our modeled gas price and expose metadata we could not legally retain publicly. I proposed content-addressed encrypted files in S3 with hashes anchored to an L2 and a Merkle root committed daily to Ethereum. I built a small proof of concept and measured a 96% reduction in projected transaction cost while preserving independently verifiable document integrity. Product approved the design after security confirmed that key rotation and access logs met the compliance requirements.”

Example answer

“Our dApp's transaction-status screen became inaccurate when our primary RPC provider degraded during a high-volume mint. I saw the issue through a jump in unknown transaction states from 0.4% to 17% and indexer lag rising above 11 minutes. I switched read traffic to a secondary provider, rate-limited status polling, and ran a reconciliation job against transaction receipts from an archive node. I declared recovery only after 99.7% of submitted transactions had a reconciled state, the indexer lag stayed below 30 seconds for an hour, and support had no unresolved payment-versus-mint discrepancies. Afterward, I added provider health scoring and an alert when pending transaction age exceeded the chain's p95 confirmation time.”

Example answer

“An auditor flagged our admin key as a centralization risk and recommended immediate removal of all upgrade authority. I agreed with the risk but disagreed that eliminating upgrades before launch was safer, because a newly deployed financial contract with no emergency remediation path could create a larger exposure. I proposed a two-of-three multisig with a 48-hour timelock, an on-chain upgrade event, and an emergency pause limited to deposits rather than withdrawals. We tested the governance flow on a Sepolia fork and measured that no single signer could upgrade, pause, or redirect funds. The auditor accepted the revised control, and we published the authority model and timelock monitoring dashboard for users.”

Example answer

“I would use an ERC-20-compatible security token with transfer hooks that check a signed compliance attestation or Merkle-based allowlist before state changes. The contract would store wallet eligibility, attestation expiry, and a hash of the policy version, while identity documents and sanctions-screening details would remain off-chain in an encrypted compliance system. Every approved or rejected transfer would emit an event that our indexer joins to the off-chain case record. Before launch, I would set a gas budget per transfer and test it against realistic allowlist sizes, targeting less than 180,000 gas on the chosen L2. In production, I would track transfer success above 99.5%, indexer-to-chain reconciliation above 99.99%, and compliance-attestation expiry failures separately from protocol failures.”

Example answer

“I begin by modeling the accounting before reading line by line: total user principal plus accrued rewards must never exceed tracked assets, and each reward interval must be claimable exactly once. I then inspect every external interaction, especially token transfers and oracle reads, to verify state updates occur before calls and that failures do not strand user balances. For reward math, I fuzz deposit, withdrawal, and claim ordering across small and large values because rounding bugs usually appear at boundaries. I use Slither for broad pattern detection, but I treat Foundry invariant tests as the decision tool because they exercise the actual accounting model. On a mainnet fork, I would also test the real token contracts and oracle feeds, then require a coverage report showing that every privileged and value-moving path has both positive and failure-case tests.”

Example answer

“I would choose Hyperledger Fabric for a multi-company supply-chain settlement network where the participants are known, pricing data must remain visible only to counterparties, and a consortium can govern certificate authorities and endorsement policies. I would use Ethereum or an L2 if token liquidity, public proof of settlement, or integration with existing DeFi rails was part of the product value. In Fabric, I would define endorsement so that both buyer and supplier organizations must approve a settlement, then benchmark the network with realistic channel and private-data workloads rather than vendor throughput claims. I would measure commit latency, failed endorsement rate, peer availability, and reconciliation exceptions between ERP records and ledger state. The decision document would make clear that Fabric improves confidentiality and governance control but gives up public composability and requires more consortium operations.”

Example answer

“For an Ethereum settlement workflow, I would distinguish submitted, included, confirmed, and finalized states in both the API and the database. A user receiving a transaction hash would see it as pending, not paid, until our watcher observed the required confirmation policy or a finalized-block signal. The event consumer would store block hash and log index, handle removed logs during a reorganization, and make downstream ledger postings idempotent. I would delay irreversible actions such as releasing custody assets until finality, while allowing low-risk UI updates after inclusion. I would monitor p50 and p95 time to finality, reorg depth, and the count of chain-to-internal-ledger mismatches, with a target of zero unreconciled finalized settlements.”

Example answer

“I would not approve a full launch of an unaudited external price feed that can influence user balances or liquidation behavior. I would offer a constrained pilot if the business deadline is real: a hard TVL cap of $100,000, a pause control governed by multisig, a maximum price-age threshold, and a circuit breaker for deviations beyond an agreed percentage. Before enabling it, I would require fork tests against the live feed, fuzz tests for stale and zero-price conditions, and an independent review focused on the new integration. I would define success as zero oracle-related reverts, zero price deviations beyond the guardrail, and no manual reconciliation exceptions during the first 14 days. If any gate failed, the feature would revert to read-only rather than exposing additional user funds.”

Example answer

“I would first disable the ownership-sync job so it could not overwrite more customer records, while keeping the raw event and job logs intact. I would group the 2,000 mismatches by contract, block range, and indexer deployment version, then compare each event against an archive-node query using block hash and log index. I would also check whether the contract changed its transfer behavior after an upgrade or whether our consumer missed removed logs during a reorg. After identifying the cause, I would backfill from canonical finalized blocks and verify the result with ownerOf calls where the token standard supports them. I would close the incident only when the mismatch count returned to zero, the backfill had a recorded checksum, and the corrected customer records were sampled against chain state at 99.99% accuracy.”

Example answer

“I would decline to add a privileged query until the request had governance approval, because private-data collection boundaries are part of the consortium's operating contract, not an application inconvenience. I would ask what reporting decision the member needs to make and determine whether monthly aggregates, anonymized benchmarks, or counterparty-approved disclosures would answer it. I would build the report from permitted summary data and document which fields remain inaccessible under the collection policy. If broader visibility were genuinely necessary, I would route a policy amendment through the consortium rather than burying an exception in chaincode. I would measure the result by report completeness, query latency, and confirmation from the affected members that no unauthorized pricing fields were exposed.”

Example answer

“I would treat this as a same-day access-control incident because a single wallet can become unavailable or compromised during offboarding. First, I would identify whether the contract uses a Transparent proxy, UUPS pattern, or AccessControl roles, then confirm exactly what the employee wallet can change. I would coordinate with security and legal to preserve authorized access long enough to execute a rehearsed role-transfer transaction to the company's multisig, ideally with a timelock if the architecture supports it. On a fork, I would test the exact transaction sequence and confirm that the multisig can perform required administrative actions without changing the implementation. I would consider the issue resolved only after on-chain role queries show the employee wallet has no authority, the multisig quorum has completed a test action, and alerts are active for every future admin event.”